Batched easing evaluation API, playlist tokenization, and Component animate refactor

[inigomonyota]

Motivation

• Reduce per-tween branches and per-frame string parsing to improve runtime locality and make the hot loop SIMD-ready by evaluating easing per-algorithm over contiguous arrays.
• Eliminate repeated playlist stringstream splitting from the hot path by parsing playlist filters once at Tween construction.

Description

• Add using EasingKernel = float (*)(float, float, float);, static EasingKernel Tween::getKernel(TweenAlgorithm), and a new batched entry static void Tween::evaluateBatch(...) which evaluates contiguous progress/start/change arrays for a single algorithm.
• Parse playlistFilter into playlistFilterTokens inside the Tween constructor and expose bool matchesPlaylist(const std::string&) const plus lightweight accessors startValue()/endValue() used by the batching flow.
• Refactor Component::animate() to a gather/bucket/evaluate/apply pipeline that: resolves start values during gather, builds a TweenEvaluation list, buckets evaluations by algorithm, assembles SoA arrays (progress, start, change) per bucket and calls Tween::evaluateBatch(...), then applies results in original order while preserving behavior for zero/negative durations and startDefined semantics.
• Add docs/animation-batching-simd-plan.md describing the motivation, design, and next steps for SIMD-friendly batching and data-layout changes.

Testing

• Ran git -C /workspace/RetroFE diff --check to validate whitespace/merge issues and it completed with no reported problems.
• Ran git -C /workspace/RetroFE status --short to confirm staged modifications and it reported the modified files.
• Inspected modified code regions with nl (slices of Tween.h, Tween.cpp, and Component.cpp) to verify the new APIs, constructor tokenization, evaluateBatch implementation, and gather/bucket/evaluate/apply pipeline; those inspections matched the intended changes.

---

Codex Task