Monotonic ULID generation performance improvements

[Seramis]

Description

• Refactored and reworked the monotonic ULID generation
  • Introduced LowLatencyLock that uses using syntax. This is faster than manual try-finally block.
  • Rewrote increment/addition logic. Now a single universal method increments by 1 or by 1-4 bytes (+1). Simpler and faster.
  • Stored last ULID and Lock struct in a separate state object that pads these fields into separate cache lines to prevent false-sharing. (Better CPU cache usage as a result)
  • Cleaned up MemoryMarshal compatibility file as some methods were not necessary anymore.
  • Performance improvement during simple +1 incrementing monotonic generation is about 10%. (from 38ns to 34ns on dev machine)
• Improved slightly parsing of ulong timestamp into bytes avoiding unnecessary span creation. Improves all ULID generation scenarios a tiny bit.
• Generalized endianness reversing throughout the implementation. Introduced Ulid.Util.cs for shared methods.
• Updated README.md with benchmarks.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update
• Other (please specify): Performance

Checklist

• The PR is submitted to the correct branch (main).
• My code follows the project's coding style. (.editorconfig)
• I have commented my code, particularly in hard-to-understand areas and public interfaces.
• I have added or updated tests for the changes I made.
• All new and existing tests passed.
• I have updated the documentation where applicable.