style: prettier format skill files

fix: detect empty-turn 200 responses as degraded, trigger fallback

Models like gemini-3.1-flash-lite under eco tier return HTTP 200 with
no content and no tool_calls (finish_reason: stop) when given complex
agentic requests (e.g. Roo Code tool schemas). This EMPTY_STREAM_ROLE_ONLY_STOP
pattern was not caught by detectDegradedSuccessResponse, so ClawRouter
treated it as success and sent the empty turn to the client.

Now detects: choices[0].message with empty content + no tool_calls +
finish_reason=stop → 'degraded response: empty turn' → triggers
fallback to next model in chain.

feat: register Predexon as OpenClaw tools + fix GET/path-param suppor…

…t in partner proxy

fix: force skill usage over browser for predexon and x-api

feat: add Predexon prediction market skill + extend partner proxy to …

…/v1/pm/*

- Add skills/predexon/SKILL.md: 39 endpoints across Polymarket, Kalshi, dFlow,
  Binance, cross-market matching, wallet analytics, smart money. Follows x-api
  format with concrete request/response examples and { data: ... } wrapper docs.
- Extend partner proxy regex to include /v1/pm/* (was only /v1/x/ and /v1/partner/)
  so ClawRouter routes prediction market calls with automatic x402 payment.
- Update skills/imagegen/SKILL.md: add image2image editing endpoint, fix size
  table with accurate dimensions per model.

fix: crash when agents.defaults.model is a string instead of object (…

…v0.12.70)

OpenClaw users with model set to plain string "blockrun/auto" caused
TypeError during plugin registration. Now auto-converts string/non-object
model values to { primary: "<value>" }.

fix: lint errors — unused import and no-useless-assignment

fix: payment settlement failure must fallback to free model

Payment settlement errors (e.g. insufficient funds on-chain) returned
non-standard HTTP codes that categorizeError() didn't recognize, so
isProviderError was false and the fallback loop broke immediately.

The user saw "Payment settlement failed" with no recovery, even though
nvidia/gpt-oss-120b (free) was available.

Fix: move payment error detection before the isProviderError gate.
When a payment error is detected on a paid model, skip directly to
the free model in the fallback chain (or append it if not present).

Reported by user Mark (chain-reaction) via logs.

chore: bump to 0.12.65 (fix free model payment error)

fix: remove unused FALLBACK_STATUS_CODES constant to fix lint