terminal: answer OSC default color queries

[amir-jakoby]

terminal: Answer OSC default color queries

Fixes #1145.
Fixes #1351.
Related to #1350, but this PR intentionally does not implement OSC 4 indexed color queries.

Adds support for OSC 10/11 default foreground/background color queries through mosh. The wrapper probes the local terminal for default colors, forwards only sanitized rgb values to mosh-server, and the server terminal emulator answers remote application queries from those values.

Description:

• Probe the local controlling terminal for OSC 10/11 replies with a short timeout.
• Forward sanitized default foreground/background colors as MOSH_DEFAULT_FG and MOSH_DEFAULT_BG.
• Answer OSC 10/11 ? queries in the server emulator only when configured values are valid.
• Add a regression test for replies, missing values, and invalid values.

Validation:

• perl -Mdiagnostics -c scripts/mosh.pl
• make -C src/tests check TESTS=terminal-osc-default-colors
• PTY wrapper simulation confirmed parsed OSC replies are forwarded into the remote server command.
• make -j4 check passed the new regression but local window-resize.test errored in this macOS/tmux environment; the log showed the driver sent qexit to the shell instead of exiting less.

[achernya]

I am not a fan of the approach using environment variables to pass this around. I am concerned about forwards and backwards compatibility.

[amir-jakoby]

@achernya I reworked this to avoid using environment variables for the wrapper/server handoff.

The revised version sends the probed default colors as internal mosh-server new -@ mosh-default-fg=... -@ mosh-default-bg=... arguments instead of MOSH_DEFAULT_FG/BG. New servers consume those -@ payloads and thread the values into the terminal object; old servers already understand -@ ARG as an argument-eating compatibility option, so they consume the pair quietly and continue without enabling the feature.

I think this is better than the env-var approach because:

• no new values leak into the user's remote shell environment or child processes;
• MOSH_DEFAULT_FG/BG do not become an accidental public ABI;
• there is no new documented/public mosh-server option surface;
• there is no protocol/protobuf change, and old-client/new-server or new-client/old-server both degrade to no OSC default-color replies;
• the terminal emulator no longer reads getenv() at query time, so the behavior is carried as server startup state instead of ambient process state.

I also kept the scope limited to OSC 10/11 default foreground/background replies and added a small regression for valid 1-4 digit XParseColor components.