Update dyload rules and version, improve test threading

[ajz34]

This pull request introduces a new release (v0.2.6) focused on improving and documenting the dynamic loading feature for the TBLIS Rust bindings. The changes enhance how the shared library is located at runtime, add detailed documentation and user guidance, and update versioning across the workspace. The most important changes are grouped below:

Dynamic Loading Improvements and Documentation:

• Expanded and clarified the dynamic loading process in tblis-ffi/src/tblis/mod.rs to search for the shared library in a prioritized sequence: user-defined environment variables, library path environment variables, Python interpreter paths, and standard system directories. Added robust Python path detection and improved error messages for easier troubleshooting. [1] [2] [3] [4] [5]
• Updated documentation in readme.md and tblis-ffi/readme.md to clearly describe the dynamic loading rules, environment variables, and how to disable/override the feature. [1] [2]

Release and Versioning:

• Bumped the workspace and crate versions to v0.2.6 in Cargo.toml, and updated dependencies accordingly.
• Added a changelog entry for v0.2.6 describing the new dynamic-loading rules. [1] [2]

Testing/CI:

• Set test threads to 1 in CI workflows for both tblis and tblis-src to ensure consistent and reliable test execution. [1] [2]

(summarized by copilot web)