Add slice and unsafeSlice for ByteArray

[jappeace-sloth]

Summary

• Adds slice :: ByteArray bs => bs -> Word -> Word -> Maybe bs — extracts sub-range [start, end), returns Nothing when indices exceed length
• Adds unsafeSlice :: ByteArray bs => bs -> Word -> Word -> bs — same but clamps out-of-bounds indices
• Both swap indices automatically if end < start
• Closes Extracting of a part #7

Test plan

• Property test: slice matches take (end-start) . drop start
• Property test: out-of-bounds indices return Nothing
• Property test: unsafeSlice clamps to valid range
• Property test: slice a b == slice b a (swap symmetry)
• All tests pass on both Bytes and ScrubbedBytes backends
• nix-build nix/ci.nix passes

🤖 Generated with Claude Code

[jappeace]

@kazu-yamamoto please review this as well, see if it's like how you expect

[kazu-yamamoto]

Just a moment.
I will try to use this function in tls and feedback my experience.

[kazu-yamamoto]

Two minor observations:

Word is difficult to use. The calling function needs to use fromIntegral extensively. Using Int and adding more guards would be more user-friendly.

The end index is not intuitive because the element indicated by start index is included but the element indicated by the end index is not.

For example, slice ba 1 1 doesn't intuitively mean 0 bytes. Also, when copying the entire array, it becomes slice ba 0 (length ba), where the end value overflows, which feels awkward.

I found the length to be more clear in meaning.

[kazu-yamamoto]

I guess that unsafeSlice should call error on error conditions so that users can notify the error quickly.

[jappeace]

@kazu-yamamoto I changed it according to your suggestions, please review if this is better for you?

[kazu-yamamoto]

Very good.
Thanks!

[kazu-yamamoto]

If you release a new version, I will also release tls.

[jappeace]

yeah it's released: https://hackage.haskell.org/package/ram
🚢