feat: add --wait flag to compute-envs delete

[robsyme]

Summary

• Adds a --wait boolean flag to tw compute-envs delete
• After the DELETE API call returns, polls describeComputeEnv() until the CE is fully removed (404) or enters ERRORED state
• Backoff: 2s initial, +1s per poll, max 120s (matches existing waitStatus pattern)
• Prints progress dots (suppressed in JSON output mode)
• Prints error message if CE enters ERRORED state

Context

Forge CE deletion is async since Dec 2025 (commit be99dcf9b9). The CE stays in DELETING status for ~2 minutes while AWS resources are cleaned up. Tools like seqerakit that delete-then-recreate CEs (on_exists: overwrite) need to wait for deletion to complete before creating the replacement.

Related: FD-7252, platform#10680

Test plan

• testDeleteWaitHappyPath: DELETE 204 -> poll DELETING -> poll 404 -> exit 0
• testDeleteWaitErrored: DELETE 204 -> poll ERRORED -> exit 1
• Existing delete tests unchanged (testDelete, testDeleteInvalidAuth, testDeleteNotFound)

🤖 Generated with Claude Code

[cristianrcv]

@robsyme To my knowledge, the Tower CLI functions primarily as a direct interface to our API, so introducing additional logic or complexity at the CLI level may not be the most appropriate approach. If an external tool requires waiting for a delete operation to complete, would it not be more suitable to implement that functionality within the tool itself?