Optimize Formatter for 90% less memory usage and 15-30% faster execution

[ThePaperPlate]

I've optimized the Formatter class to significantly improve its memory efficiency and performance

Main Improvements

1. Added Object Pooling: Implemented a custom Poolable<T> system that recycles Element, Token, Lexer, Entry, and ElementTreeBuilder objects
2. Replaced the delegate-based Lexer state machine with a more efficient enum-based implementation
3. Optimized String Handling: Implemented StringBuilder pooling within ToTreeFormat()
4. Cached Common Objects: Created static cached instances for frequently used objects like translation dictionaries and common tags
5. Maintained API Compatibility: Preserved the original public APIs so third-parties wouldn't need modifications

Benchmark Results

Using this crude benchmark script on both the original and optimized versions:

Before vs After (100 iterations on 2053 test inputs):

Execution Time (lower is better)

Memory Usage (lower is better)

Formatter	Original Time	Optimized Time	Time Improvement	Original Memory	Optimized Memory	Memory Reduction
ToPlaintext	3.41 ms	2.91 ms	14.7% faster	1.59 GB	158.23 MB	90.3% less
ToUnity	2.34 ms	1.61 ms	31.2% faster	1.76 GB	185.40 MB	89.7% less
ToRustLegacy	1.85 ms	1.51 ms	18.4% faster	1.66 GB	178.11 MB	89.5% less
ToRoKAnd7DTD	1.85 ms	1.51 ms	18.4% faster	1.65 GB	173.36 MB	89.7% less
ToTerraria	1.84 ms	1.51 ms	17.9% faster	1.64 GB	173.28 MB	89.7% less
Parse	1.46 ms	1.19 ms	18.5% faster	1.36 GB	254.18 MB	81.7% less