Improve command-buffer bake efficiency and add inspect chain helper

[mqxf]

Motivation

• Reduce unnecessary allocations and unsafe mut-casting in command-buffer baking to make finish faster and safer.
• Provide a small, useful chainable primitive so users can run side effects (logging/metrics/validation) inside command chains without breaking the chain.

Description

• Replaced drain(..).collect() uses for extracting tasks/controllers with std::mem::take to avoid extra allocations and drains in TaskChain::chain_semaphores and TaskChain::get_controllers (src/command_buffers/buffer.rs).
• Changed RawCommandBuffer::finish to take &mut self, removed the internal unsafe self-mutation, precomputed the final task capacity, and incrementally extend the output vector to avoid an extra intermediate collect step (src/command_buffers/buffer.rs).
• Added a new extension trait Inspect<T> with a chainable method .inspect(...) that runs a side-effect taking &T and returns the original value unchanged, enabling non-intrusive inspection in command chains (src/command_buffers/commands.rs).
• Minor formatting and ergonomic cleanups to signatures and whitespace (small refactors across buffer.rs/commands.rs).
• Public command-buffer API unchanged; improvements are internal and additive.

Testing

• Ran cargo test --all-features, which completed successfully (unit tests and doc-tests passed).

---

Codex Task