feat(s3_bucket_downloader): add S3 bucket downloader tool

[sangshuduo]

• Implemented a new tool to download files from an S3 bucket using multiple threads.
• Added command-line arguments for bucket name, output directory, number of workers, and retries.
• Included a feature to read file lists from a specified file or list objects directly from the bucket.
• Utilized libraries such as aws-sdk-s3 and tokio for asynchronous operations.

[mentatbot]

I'm getting to work. See my steps and track my cost here.

• Wake on any new activity.
  _{I always see new activity while actively working. When paused, I'll only respond when tagged unless this box is checked.}

[mentatbot]

I've taken a look at your S3 bucket downloader implementation. This looks like a great addition to the toolkit that complements the existing random_pairs_of_s3file tool nicely!

A few thoughts:

Strengths:

• Great use of multi-threading with rayon to parallelize downloads
• The progress bars are well-implemented and provide good user feedback
• Smart caching of file lists to avoid unnecessary bucket listing operations
• Solid error handling with retries and exponential backoff

Potential improvements:

1. The code uses div_ceil which might require a newer Rust version - we should verify this works across all target Rust versions in CI.
2. Loading entire files into memory with Vec<u8> could be problematic for very large files - maybe consider a streaming approach for large files.
3. It might be helpful to add some documentation in the README.md about this tool, similar to what was done for random_pairs_of_s3file.

Let's see what the CI says, but this looks like a solid implementation to me!