Allow disabling x86 emulators and/or automatically disable them based on host arch

[nrabulinski]

Muvm can be useful on x86 systems too, maybe not for page size mismatches (though, maybe?), but e.g. for isolation. On such systems it doesn't make sense that muvm tries to configure a binfmt handler.

My idea was to allow setting the --emu argument to none, as well as automatically set it to none instead of fex when muvm is compiled for x86. Would that be appreciated?

This also ties into what I mentioned in #158 with a config file where you can define custom emulators, if you're running on some other "exotic" architecture and want to utilize muvm for page size compatibility.

[sbrivio-rh]

My idea was to allow setting the --emu argument to none, as well as automatically set it to none instead of fex when muvm is compiled for x86. Would that be appreciated?

Absolutely, yes. I'd call that "needed". :)

Muvm can be useful on x86 systems too, maybe not for page size mismatches (though, maybe?), but e.g. for isolation. On such systems it doesn't make sense that muvm tries to configure a binfmt handler.

By the way, my use case would be testing kernel changes, and possibly passt changes, especially in combination with containers/libkrun#256 (right now the workaround is rather awkward so it's not part of my workflow).

The whole thing should also make running kselftests much faster and convenient. I started mbuto for that a long time ago, but when it comes to Python dependencies, or testing nft (or passt) changes, that approach is not exactly convenient. Having the current filesystem seamlessly exported to the guest makes it simpler.