feat(dot): add --lock-wait for a running history operation

[zachbornheimer]

GitHub will not let me open a pull request on this repository, so the patch is on a branch:

https://github.com/zachbornheimer/mise/tree/dot-lock-wait

Compare: main...zachbornheimer:mise:dot-lock-wait

Motivation

A mutating mise dot command already waits 30 seconds for the history operation lock, then fails with another history operation is running. A script that needs its own checkpoint, with a description, label, or task id, has to retry in a shell loop. The watcher will record an unlabeled change, but it will not record the checkpoint the command asked for.

Changes

mise dot --lock-wait <duration> and MISE_HISTORY_LOCK_WAIT set that wait for this process. 0 fails at once. Longer than 1 hour is refused. The wait stays inside acquire_operation_lock, so the command takes the lock itself instead of polling and then racing the watcher.

• crates/mise-system/src/history/scope.rs — operation_lock_wait() replaces the 30-second constant in begin_kind, take_operation_lock, and the late journal open. Callers that pass zero (begin_automatic_apply, recovery_lock) do not read it.
• src/cli/dotfiles/mod.rs — the flag. watch and recover ignore it and warn, so an exported MISE_HISTORY_LOCK_WAIT does not stall the watcher.
• docs/dotfiles/reference.md — the history lock section.

Not included: a bare --wait that means forever, and making the watcher or recovery wait.

Testing

• cargo test -p mise -p mise-system --lib configured_lock_wait_is_honored_and_zero_callers_are_not — the override waits until the holder releases, 0 fails at once, and recovery_lock still does not wait. Passed.
• cargo check -p mise --bin mise — the flag compiles. Passed.

Untested: the generated CLI pages. mise usage did not finish locally, so docs/cli/ and mise.usage.kdl are unchanged.

[github-actions[bot]]

Implemented in #15498: feat(dot): add history.lock_wait to wait longer for a running history operation