If Apple didn't care about gaming they would not invest tons of $$$ and time to design GPUs with unique gaming only features... Apple cares a lot about gaming, and the last few years of their hardware and software roadmap makes it very clear.
There is gaming and then there is gaming. What we call "hardcore gamers" are definitely not Apple's target demographics. But home users who want to play games are.
To put to in the context of your argument, no, if you only care about gaming, there are definitely better options out there. But if you care about having a Mac (be it for work or personal computing), you should be able to use it for gaming if you so desire.
Or to put it even simpler: Apple does not care about giving you gaming super-GPUs. What they care about is putting a GPU that's good enough for gaming inside every Mac.
It doesn't have to be an exclusive, games can target multiple rendering APIs. Or just use Vulkan over a Metal translation layer. Check out the recently released Metro: Last Light or Baldur's Gates 3 (which will probably be first Apple Silicon native AAA game).
Of course it is. Apple's M1 is roughly comparable in performance to a 50-60W Nvidia Pascal GPU or a 20-30W Turing. In a game that takes full advantage of Apple's unique GPU features, M1 is basically equivalent to a GT 1650.
It's the question of supply of demand. I completely agree with you that it's probably not worth it for a regular gamedev today to use Metal directly — unless they design their game from the start to be cross-platform, in which case adding direct Metal support is fairly easy. Luckily for us Mac users, most games use middleware engines that support Metal (or they use Vulkan which again can be easily utilized on macOS via MoltenVK).
As Apple Silicon market share grows and the number of people playing games increases, it will become an economic signal for the game industry, incentivizing developers to improve support and performance on Macs. That is when we will see more direct Metal support coming in. But of course, it will still take a while until Mac is seen as a platform of serious interest for game developers.
And of course, Apple Silicon Macs have a big advantage for game developers: they are a unified hardware platform with console-like level of hardware access. You don't have to worry about which features are fast or not or about some GPU-specific behavior, as all of them share the same hardware architecture. This radically simplifies engine development and reduces the amount of testing you have to do.
Because a) Vulkan is extremely complicated to use and Apple wanted a user-friendly API and b) Vulkan has to support a wide range of hardware, so it targets a hypothetical "common GPU architecture". Apple GPU have a lot of unique features that would require extensive Apple-specific Vulkan extensions to support properly.
Metal allows Apple to rapidly innovate in the hardware space without fighting the API, while offering a low-level software interface that exposes hardware details and makes it easy to utilize the fast path. Besides, there is the question of API usability. I have worked with most GPU APIs and I believe that Metal is hands down the best GPU API that currently exists. It's very easy to learn and use, extremely well designed, and very very powerful (especially on Apple hardware).
That said, MoltenVK works great on Apple. You can still get more performance out by using Metal directly and rewriting your engine to use unique Apple features, but as you say yourself it is more work and not always viable.