test(pty): wait for the line editor instead of a fixed 600 ms (closes #58)

[ShortArrow]

Closes #58.

Why the bootstrap waits for a signal instead of a fixed time

PtySession::spawn used to drain the shell's output for a fixed 600 ms before sending the first bootstrap line. pwsh's PSReadLine starts reading keys only after it has sent its cursor-position query, about 300 ms after launch on an idle machine. On a slow macOS runner it started after the 600 ms had passed, so the first line was rendered but never run, and the step timed out after 5 s. With the failure output from #60, the CI failure showed exactly that: the whole line drawn, then no newline.

The harness now waits until each shell's line editor reads keys. pwsh sends two cursor-position queries, nu's reedline sends one, and bash and zsh send none and print a prompt. Each editor blocks until its query is answered, so the queries arrive one at a time and can be counted. After the signal the shell must print nothing for 100 ms. A shell that is not ready within 5 s fails at the startup step.

How it was checked

With 96 busy processes on a 32-core WSL machine, the pwsh PTY tests failed 30 of 30 runs with the fixed wait, every time at the same step as on macOS CI, and 0 of 30 runs with this change. Under the same load the bash, zsh, nu and pwsh tests each passed 10 of 10 runs.