Retry websocket connections closed with code 1001

[patrickswedish]

Closes #63.

What changed

• treat a remote WebSocket close with status 1001 (Going Away) as a recoverable disconnect
• invoke on_close once for that remote lifecycle event
• reuse the existing reconnect and resubscribe path when retries are enabled
• preserve deliberate client shutdown as terminal behavior
• extend the mock WebSocket server with a real 1001 restart scenario

Root cause

websockets classifies both 1000 and 1001 as ConnectionClosedOK. The message handler previously exited for every ConnectionClosedOK, so a Coinbase service restart silently stopped the receive loop without notifying the caller or attempting to restore subscriptions.

Validation

• python -m unittest tests.websocket.test_websocket_base — 20 tests passed
• python -m unittest discover -v — 174 tests passed
• black --check passes for the three changed files after formatting
• git diff --check passes

The new regression test uses the local WebSocket server to send status 1001, restart, and verify both the close callback and subscription restoration.

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 1 Sum 2

[patrickswedish]

Checking in on this — happy to address any review feedback or rebase if needed. All 174 tests pass and the change is scoped to the 1001 close-code handling path with no impact on normal shutdown behavior.