I don't think that it is terribly unusual to have historical settings. And no, this is not an issue for me in practice, as any sensible player would understand that their ancient Egyptian doesn't know how to make gunpoweder and thus would never attempt it.
Let's go here, then. Having a historical setting can be lots of fun, but the thing is that nothing in you historical setting game actually ever happened -- it's all made up. If so, why can you not make up someone in Ancient Egypt inventing gunpowder? It might even have happened -- we already have strong historical proof that gunpowder existed in China and pretty much nothing was done with it, so couldn't a lone inventor create gunpowder in ancient Egypt and then just not do anything with it? Having created gunpowder does bring on firearms or the modern age, after all. So, already, your premise that a historical setting prevents such things is on poor footing.
The only reason I can think to restrict a historical setting like this would be because that's not the story the GM wants to tell or have told in game. If so, perhaps the GM needs to be more forthright in setting expectations? If they aren't, any knockon problems caused by having a setting where gunpowder can exist but then becoming mad if anyone tries to make it is, again, the GM's problem. Simply solved with simple solutions.
Nothing you say here makes any sense. In your paradigm GM cannot deny actions that are physically possible such as mixing these ingredients. And only in your paradigm whether the GM knows the formula matter because in mine we won't ever get that far.
You're 100% right -- the GM should not deny the attempt to do something. However, that doesn't mean the GM is incapable of adjudicating that attempted action. Let's say that a PC tries to mix the ingredients for gunpowder. Assuming it's possible in the game setting, they'd need to have these ingredients, have purified them properly, mix them in the proper proportions, and then maybe it works. I can challenge this at the point of even having the ingredients -- that looks like multiple quests to me to locate good sources, each of which with a strong chance of failure. I can challenge this at the point of refinement -- that requires the proper tools and methods which, even if known to the player who recites them, doesn't mean their character performs successfully, so lots of chances for failure. Finally, we get to the part you mention -- mixing. The player can recite the proportions, but that doesn't mean the character successfully accomplishes this. Look like more chances for failure.
And, all of that assumes that I, as GM, rule these things as uncertain. I could rule them as automatic successes. I could rule any as an automatic failure. This is my authority -- the authority to adjudicate actions. I don't need to police the PC's thoughts -- I have infinite dragons for cripes sake, I can let the player have their PC's thoughts.
Because knowledge is part of the fictional reality. And funny thing about 'you don't know it's an orc.' I have literally played this exact thing. I were playing in a campaign where the setting was such that at least initially it seemed that the humans were the only sapient species that existed. Early on we learned about existence of demons. But at one point we encountered these tusked green creatures. They didn't speak our language and we never learned what they were called. So never in-character did we call them orcs, because the characters obviously couldn't have known that.
Of course knowledge is part of the fictional reality -- the GM determines what's true. But knowledge isn't the PC's thinking. The PC's thinking, and what they say, and the actions they attempt are all not knowledge. Anything a PC says is only true if the GM determines it to be so.
But, let's put that aside. Let's say that, for whatever reason, the GM has determined thing A to be true. The player, though reading something, also knows thing A. In play, without establishing prior that the PC also knows thing A, the player declares their PC speaks thing A. This is the core case for "metagaming," right? Here's the thing, the PC thinking thing A doesn't mean it's true or anyone believes them, nor does it render thing A moot. As the OP went, the situation went from "this NPC might secretly betray the PCs because she's a secret lich" to "this NPC now knows the PCs know about her secret lichdom, and is planning to betray them to keep the secret." Story continued. The only reason to be upset about this is if thing A means the GM can no longer tell the planned story the way the GM planned it. This forms the core of the play aesthetic argument as well, if removed by a step.
As for you pretending you didn't know what orcs were, cool, glad you had fun. I'm not going to interfere if players want to have that fun, but I'm also not going to care if they don't -- the challenges I present aren't going to depend on the players pretending to not know what orcs are. It's entirely up to the players if they want to engage in that, not me -- I'm not going to force players to pretend they don't know what an orc is. This seems like a waste of time.
Your category error is this arbitrary distinction between physical and mental. Both can be just information about the fictional reality. Your character can't reach that' and 'your character don't know that' are not fundamentally different types of statements, they both clarify to the player the reality in which the character exists.
No, it's the difference between what a PC thinks and what's true. My job as GM is to make sure everyone is on the same page as for what's been established in the scene, not what a PC thinks about what's established in a scene. There's absolutely no distinction here between physical and mental, that's just your misunderstanding.
Look, knowledge skills are extremely valuable in my game and used all the time. I also don't care about metagaming. You really need to make an effort to hold both of these thoughts in your head at the same time and try to find a way to understand how both can be true. What you should stop doing is repeatedly telling me I must not mean one or the other because you can't grasp both at the same time.
Majority of tabletop RPGs follow my assumptions here. There are some though which give the players some meta-level control. In such games it is pretty much always explicitly spelled out and follows some defined mechanics.
They don't, actually. Most don't mention metagaming at all, or do so in way like 5e does -- not at all about PC vs player knowledge. You can trivially prove me wrong by presenting the rules quotes from these games. Already, 5e doesn't agree with you as it has no explicit statements to support you, even while making explicit statements about metagaming that do not support your theory. Note -- you can absolutely have an anti-"metagaming" table rule in 5e, and it does encourage you to consider your table rules, so there's support for your play preferences in that, but that's not following your assumption, it's allowing for it. That's the industry leader out of the way. Powered by the Apocalypse games are out, as is FATE. That's some other industry leaders. Which of the remaining games makes explicit statements that follow your assumptions, again?
