Support for o200k_base and gpt-4o (omni) model

[winzig]

Notes:

• I added in the new test plans for o200k based on output from tiktoken
• I added a SetupO200k() in CompareBenchmark (and renamed Setup to SetupCL200k), but it's not flagged as the GlobalSetup, so not currently used/benchmarked. Wasn't sure if you wanted to re-benchmark with the new encoding?

[winzig]

Fixed some issues that I found, all tests now passing.

[dmitry-brazhenko]

Thank you so much for contribution!

[dmitry-brazhenko]

I added minor fixes around failed tests in macos pipeline and will merge it here: https://github.com/dmitry-brazhenko/SharpToken/pull/43

[dmitry-brazhenko]

Did it here: https://github.com/dmitry-brazhenko/SharpToken/pull/43