Add BUILDKITE_GIT_CHECKOUT_TIMEOUT timeout to add upper duration on the checkout hook

[brandonnovick]

Description

Adds a new --git-checkout-timeout flag / BUILDKITE_GIT_CHECKOUT_TIMEOUT environment variable that allows configuring a timeout (in seconds) for the entire git checkout phase. When the timeout is exceeded, the job fails with exit code 95. When unset or set to 0, checkout runs with no timeout (existing behavior).

Context

Long-running or hung git operations (clone, fetch, checkout, submodule update) can cause jobs to hang indefinitely. This is particularly bad in this modern era of GitHub where conditions have been suffering recently. This provides a mechanism to fail fast with a clear exit code when checkout takes too long. This is useful for CI environments where retries are leaned upon heavily , even in the face of needed infrastructure improvements (potential network throttling, etc. )

Changes

• Added GitCheckoutTimeout config field to ExecutorConfig with
  env:"BUILDKITE_GIT_CHECKOUT_TIMEOUT" tag
• Added --git-checkout-timeout CLI flag (default 0, meaning no timeout) in global.go
• Wired the flag through BootstrapConfig, AgentStartConfig, and AgentConfiguration
• Added BUILDKITE_GIT_CHECKOUT_TIMEOUT to the protected env vars list
• Added checkoutWithTimeout() wrapper in checkout.go that applies context.WithTimeout around
  the checkout call; returns shell.ExitError{Code: 95} on timeout
• Job runner sets BUILDKITE_GIT_CHECKOUT_TIMEOUT env var for child processes when configured
• Added reflect.Int support to ExecutorConfig.ReadFromEnvironment so int-typed env-tagged
  fields are loaded correctly

Testing

• Tests have run locally (with go test ./...). Buildkite employees may check this if the
  pipeline has run automatically.
• Code is formatted (with go tool gofumpt -extra -w .)

Contributions

This PR was largely generated by Claude code, but reviewed manually and tweaked by a human

[zhming0]

Thank you so much for raising this PR @brandonnovick , we agree with the general direction but we think it's best if we be putting the timeout per attempt.

Please let us know you want to give it an another shot, or alternatively raising a feature request to us.

Lastly, for our curiosity, how often is such git hanging issue happening for you?

[zhming0]

Adding a global timeout to checkout is something that we are a bit hesitant about, given there is an retry mechanism underneath the checkout.

However, we think it's very reasonable to add a timeout setting for each retry attempt, it would:

• Kill a hung individual git operation quickly
• Still allow the retry loop to kick in and try again

Unblocking CI is our ultimate goal.