Fix docs_from_attrs truncating mid-UTF-8 codepoint

[leighmcculloch]

What

Use floor_char_boundary to truncate doc comments at a valid UTF-8 character boundary in docs_from_attrs. Add a test that confirms a doc string where a multi-byte character straddles the truncation boundary produces valid UTF-8.

Why

The current implementation truncates doc bytes with Vec::truncate at an arbitrary byte offset, which could split multi-byte UTF-8 codepoints and store invalid UTF-8 in the contract's spec XDR.

Close #1768

[Copilot]

Pull request overview

This PR fixes a UTF-8 correctness bug in soroban-sdk-macros where doc-comment truncation could split multi-byte characters, potentially encoding invalid UTF-8 into the contract spec XDR.

Changes:

• Truncate doc strings using floor_char_boundary so truncation always occurs at a valid UTF-8 boundary.
• Add a unit test covering the case where a multi-byte character straddles the truncation boundary.