Add WordCount for mnemonic generation

[coreyphillips]

Adds a public WordCount enum and updates generate, generate_in, and generate_in_with to accept it instead of usize. The enum restricts generation to the five BIP39 word counts.

This is a breaking API change. Callers must replace integer word counts with the corresponding WordCount variant.

Closes #97.

AI tooling was used to help revise this change.

[caarloshenriq]

tACK

I tested the changes locally:

• Ran the full test suite (cargo test) and all tests passed.
• Performed manual tests by generating mnemonics with different WordCount variants (12, 18, and 24 words) using generate, generate_in, and generate_in_with.
• All generated mnemonics had the correct word count and the generation flow behaved as expected.

Everything worked as intended.

[tnull]

Thanks for looking into this! While it's a nice API improvement, it's probably not super critical so I think we want to hold off on merging this until we have another, more critical change that would require breaking API anyways.

In the meantime, here are a few comments. Please also always disclose in the PR description and commit message if you utilized AI tooling. To this end, rewriting the PR description and commit messages to replace AI slop with concise descriptions would be appreciated. Thanks!

[coreyphillips]

Thanks. I addressed the inline comments, updated the benchmark call site, and rewrote the PR description and commit message with an AI tooling disclosure.