acp_thread: Forward token usage from PromptResponse to thread UI

[hahahuy]

Fixes #53130

Summary

External ACP agents (e.g. Claude Code CLI) return token usage in PromptResponse via the unstable_session_usage feature, which Zed already enables. Previously this data was discarded — token_usage stayed None on the AcpThread, so render_token_usage() returned early and showed nothing.

Changes:

• acp_thread.rs: Extract r.usage in the Ok(r) branch of run_turn and call update_token_usage() each turn.
• thread_view.rs: Add a max_tokens == 0 branch in render_token_usage — since the ACP protocol does not expose context window size, render a plain "N tokens" label instead of a circular progress ring (which would show meaningless "0% • N / 0").
• Two new #[gpui::test] tests: one asserting usage is populated after a turn with usage data, one asserting token_usage stays None when PromptResponse has no usage.

Test plan

• Connect an external ACP agent (e.g. claude CLI) and send a message
• Confirm a token count label (e.g. "1.5k tokens") appears in the thread header after the first response
• Confirm built-in Zed agent token display (circular ring with percentage) is unchanged
• Run cargo test -p acp_thread test_token_usage

Release Notes:

• Fixed missing token usage display for external ACP agents (e.g. Claude Code CLI) in the agent panel

[cla-bot]

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: hahuy.
This is most likely caused by a git client misconfiguration; please make sure to:

1. check if your git client is configured with an email to sign commits git config --list | grep email
2. If not, set it up using git config --global user.email email@example.com
3. Make sure that the git commit email is configured in your GitHub account settings, see https://github.com/settings/emails

[cla-bot]

We require contributors to sign our Contributor License Agreement, and we don't have @hahahuy on file. You can sign our CLA at https://zed.dev/cla. Once you've signed, post a comment here that says '@cla-bot check'.

[hahahuy]

@cla-bot check

[cla-bot]

We require contributors to sign our Contributor License Agreement, and we don't have @hahahuy on file. You can sign our CLA at https://zed.dev/cla. Once you've signed, post a comment here that says '@cla-bot check'.

[cla-bot]

The cla-bot has been summoned, and re-checked this pull request!

[hahahuy]

@cla-bot check

[cla-bot]

We require contributors to sign our Contributor License Agreement, and we don't have @hahahuy on file. You can sign our CLA at https://zed.dev/cla. Once you've signed, post a comment here that says '@cla-bot check'.

[cla-bot]

The cla-bot has been summoned, and re-checked this pull request!

[hahahuy]

@cla-bot check

[cla-bot]

The cla-bot has been summoned, and re-checked this pull request!

[hahahuy]

Inspired by my Claude CLI's terminal output, which displays per-turn token usage, context window consumption, model name, and relevant context percentage:

Before:

After:

A small change — but knowing you're at 30K/200K context without leaving the editor is genuinely useful when working with long conversations.

Once this lands, I plan to extend it with the remaining indicators:

• Tokens consumed this turn
• Context window fill (e.g. 30K / 200K)
• Model name
• Relevant context %