stty: run the tests that were ignored for lack of a tty

[harshasiddartha]

Thirty-one tests in tests/by-util/test_stty.rs were marked #[ignore = "Fails because cargo test does not run in a tty"], so they never ran. They did not actually need the harness to be attached to a terminal: they either used stdin, or called terminal_simulation(true), which gives every new_ucmd!() invocation its own pty, so a setting applied by one invocation could never be seen by the --all read-back in the next one.

They now share a single pty via pty_path() and --file, the way the tests in the same file that already run do. A few assertions were written against output stty does not produce and are corrected: size prints <rows> <columns>, and settings that equal the sane defaults only show up under --all.

Running them uncovered a real bug. cs5/cs6/cs7 are declared as members of the CSIZE group, but the bit was OR'd into the control flags without clearing the group first, so stty cs7 on a terminal already set to cs8 exited 0 and left it at cs8. A flag's group is now cleared before the flag is set.

The read-back assertions compare whole whitespace-delimited tokens rather than substrings. stdout_contains("parenb") also matches -parenb, so those checks passed whether or not the setting had been applied; the same held for icanon and opost. Flipping a setting in one of the converted tests now fails it, where the substring form passed.

On macOS the file goes from 51 passing / 32 ignored to 81 passing / 2 ignored. speed is not implemented in this stty, so test_print_speed stays ignored, now with a reason that says so, and special_settings asserts the operand's current rejection so it is still covered.

Fixes #14370

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/cut/bounded-memory (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/misc/tty-eof (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/tail/truncate is no longer failing!