fix(anthropic): warn when response is truncated by max_tokens limit

[citizen204]

Summary

Closes #1982

When the Anthropic API stops generation because the response hit the max_tokens limit, users currently receive silently truncated output with no indication of what happened. This PR adds a WARNING-level log message in both the streaming and non-streaming paths so the problem is immediately visible.

• In try_run: checks response.stop_reason == "max_tokens" and logs a warning with the current max_tokens value.
• In try_stream: checks event.delta.stop_reason == "max_tokens" in the message_delta event and logs the same warning.

No behaviour changes — the message is still returned as-is; users who want longer output can simply increase max_tokens.

Test plan

• test_try_run_warns_on_max_tokens — mocks a response with stop_reason="max_tokens", asserts a WARNING containing "max_tokens" is emitted.
• test_try_stream_warns_on_max_tokens — mocks a stream with a message_delta whose delta.stop_reason="max_tokens", asserts same warning.
• Existing test_try_run / test_try_stream_run tests are unaffected (mock responses have a non-string stop_reason that never equals "max_tokens").

---

📚 Documentation preview 📚: https://griptape--2212.org.readthedocs.build//2212/

[collindutter]

@citizen204 this was made against the dev branch, please make the PR against the main branch.

[citizen204]

Hi @collindutter — just to flag: the PR is currently targeting main (you can verify in the PR header). Happy to address any other concerns you have! Let me know if there's anything else to adjust.