Honor ConsumerConfig.durable_name in pull_subscribe

[caspervonb]

When the durable was supplied only through the config, pull_subscribe generated an ephemeral name while leaving durable_name set, so the server rejected the mismatched request. Resolve the durable from config.durable_name before the lookup.

Fixes #603.

[synadia-claude-reviewer]

Claude finished @caspervonb's task in 1m 27s —— View job

---

Code Review

• Gather context (reading PR diff and source files)
• Analyze changes for correctness, security, and performance
• Post inline comments on specific issues
• Summarize findings

---

The fix is correct and minimal. The root cause was clear: pull_subscribe never consulted config.durable_name when the durable positional argument was omitted, so it fell through to generate an ephemeral consumer_name at line 611, producing a name mismatch that the server rejected.

The placement of the new check (after stream lookup, before the should_create block) is exactly right, and config.durable_name as a truthiness check safely covers both None and empty string.

One minor gap in the test: it only verifies the consumer exists after a single subscribe call. It doesn't verify that calling pull_subscribe a second time with the same config re-binds to the existing consumer (i.e. should_create = False) rather than erroring or creating a second one. This is the behavior most likely to regress silently, since the whole point of a durable is idempotent reuse. Worth adding a second pull_subscribe call and asserting the same consumer info is returned.

No correctness bugs, no security or async pattern issues. The fix is ready to merge as-is; the second test case would be a nice addition but is not blocking.