v0.5: Hardening, edge cases and reconnect policy

[Mewt0]

Goal

Harden SmartReconnect against races, unusual battle states and long outages before release.

Scope

• Explicit reconnect state machine and transition logging.
• Cooldown and retry/backoff policy.
• Handle client already disconnected before SmartReconnect acts.
• Handle battle exit, arena change, postmortem/spectator, prebattle/loading and afterbattle states.
• Handle long local internet outage and server outage/restart.
• Ensure callbacks/observers/events are always removed on fini/transition.
• Confirm no reconnect loops after failed WGC login.
• Review grace period using real v0.1 telemetry rather than guessing.

Acceptance criteria

• No duplicate callbacks or observers after repeated battles/reconnects.
• No automatic action in replay.
• No infinite reconnect/login loop.
• One outage is bounded by configured attempt/cooldown policy.
• Every terminal path returns the controller to a known safe state.
• python.log is sufficient to reconstruct the state transitions of a failed session.

Test matrix

Include threshold-boundary race, client timeout race, battle ending during reconnect, reconnect after death/spectator, server unavailable, token failure, multiple successive outages, and clean client shutdown.