[BUG] — `AiTokenLimiterPlugin`: `chunk.split("\\r?\\n")` (3-char regex, no fast path) per SSE chunk

- severity: Medium
- files: `shenyu-plugin-ai/shenyu-plugin-ai-token-limiter/.../AiTokenLimiterPlugin.java:268` (inside `appendResponse`'s `doOnNext`, per data buffer chunk of a streaming response)
- description: `String.split("\\r?\\n")` runs `Pattern.compile` per chunk; SSE streams emit many chunks per response. (Sibling `COMPLETION_TOKENS_PATTERN` at `:79` and `MAPPER` at `:196` are correctly precompiled.)
- impact: Per-chunk regex compilation on streaming AI responses.
- suggested_fix: Precompile `private static final Pattern SSE_LINE = Pattern.compile("\\r?\\n");`.
- confidence: High

---
_Identified during the 2026-08-02 deep re-scan; full list in [`docs/scan2-2026-08-02/06-medium-tiers.md`](docs/scan2-2026-08-02/06-medium-tiers.md)._

Read shenyu-plugin-ai/shenyu-plugin-ai-token-limiter/.../AiTokenLimiterPlugin.java, starting at appendResponse and line 268 in the doOnNext handler. Compare the existing precompiled COMPLETION_TOKENS_PATTERN and MAPPER fields, then verify that SSE chunks no longer trigger regex compilation per chunk.

Written by the indexing model from the issue text.