Add --name compatibility for local server teardown

[sdairs]

Summary

• accept canonical positional names and compatibility --name for local server stop and remove
• preserve omitted names through clap and dispatch while retaining the current default runtime fallback
• advertise only positional names in agent-context help so the compatibility flag can age out
• cover conflicts, omission, trailing options, exact dispatch identity, and agent-help guidance

Tests

• cargo test -p clickhousectl
• cargo fmt --all --check
• cargo clippy -p clickhousectl --all-targets -- -D warnings

Closes #474

[sdairs]

Carry-over review from PR #488 (old PR for #474). The clap shape is identical to #488 and the subprocess tests here are genuinely better (decoy-server dispatch-identity assertions, no-side-effect conflict check, dedicated tests/local_server_name_compatibility_test.rs). But this PR directly violates the one substantive sdairs review comment on #488. Required:

1. Stop advertising --name in the agent-context help. On Add --name compatibility to local server stop and remove #488, sdairs wrote: "remove the added mention of --name in the agent context help. the flag exists as compat but should not be advertised to aide in aging it out." Add --name compatibility to local server stop and remove #488 fixed this in commit fc70ebf ("Stop advertising teardown name flag"); its final help only says "pass it positionally to select another server (e.g., server stop dev)". In this PR, both agent-context blocks advertise the compat form:

   • crates/clickhousectl/src/local/cli.rs:~421-422 (stop): "The compatibility form --name dev remains accepted, but cannot be combined with a positional name."
   • crates/clickhousectl/src/local/cli.rs:~467-468 (remove): same sentence.
     Delete both sentences (port the intent of fc70ebf). The /// Compatibility form for the server name; prefer positional NAME flag doc lines are milder (Add --name compatibility to local server stop and remove #488's final state kept similar), but combined with the above the compat form is heavily advertised — reconsider those too.

2. Consider whether the stop_server()/remove_server() extraction in mod.rs (~70 lines of churn) is needed — Add --name compatibility to local server stop and remove #488 proved two-line inline resolution suffices. Defensible tidiness, but borderline for a compat-flag PR.

3. CI note: local postgres edge cases fails here but identically across the stack base — inherited from Validate local Postgres start options #547's script drift, not this PR's change.