feat(wasix): report stdin/out/err as Unknown if it isnt a tty

[kilyanni]

This PR makes stdin/out/err report whether they're a tty or a file

Apps that check whether their IO is interactive need this in order to function properly

[Copilot]

Pull request overview

This PR updates WASIX fdstat reporting for the standard file descriptors (stdin/stdout/stderr) so that applications can distinguish interactive (TTY) I/O from non-interactive I/O.

Changes:

• Added a helper to map “is this stdio a TTY?” into a WASI Filetype.
• Updated fdstat() for fds 0/1/2 to return CharacterDevice only when the corresponding host stream is a TTY, otherwise Unknown.

Suppressed comments (2)

lib/wasix/src/fs/mod.rs:1865

• The TTY detection for stdout is based on the host process’ std::io::stdout(), but fd 1 can be overridden via WasiFs::swap_file. When stdout is swapped to a non-host stream, fdstat(1) should not consult the host TTY status.

                    fs_filetype: Self::std_fd_filetype(std::io::IsTerminal::is_terminal(
                        &std::io::stdout(),
                    )),

lib/wasix/src/fs/mod.rs:1875

• The TTY detection for stderr is based on the host process’ std::io::stderr(), but fd 2 can be overridden via WasiFs::swap_file. When stderr is swapped to a non-host stream, fdstat(2) should not consult the host TTY status.

                    fs_filetype: Self::std_fd_filetype(std::io::IsTerminal::is_terminal(
                        &std::io::stderr(),
                    )),

[Arshia001]

Having to change the wasi_fyi suite is a signal we're doing something wrong, since this test suite was created against the WASIp1 standard rather than our implementation of WASIX.

[kilyanni]

Hm. The test suite asserts unconditionally that std* should be ttys. In our test-harness this is very likely not the case, so I'd expect this to be changed now.

[Arshia001]

I think this breaks anything that's not directly connected to a host's TTY, including something like wasmer.sh's terminal emulation.

[kilyanni]

But generally yeah, we should probably have some way to 'lie' to consumers that std* are TTYs even if they aren't with some (maybe on by default?) flag.

Maybe --tty=on|off|auto?