Use byte widths for fixed-record MIME databases

[OskarEichler]

Summary

Measure database row sizes and generated column padding in bytes. Positional reads use byte offsets, but String#length counts characters; custom Unicode extensions could make later records unreadable.

Reproduction

Write a UTF-8 custom extension database with these equal-byte-width rows:

é text/one 7bit
ê text/two 7bit

RandomAccessDb.new(path, 0).lookup('ê') returned nil. It now returns text/two. Three focused lookup/miss assertions pass with native pread and again with the fallback seek/read implementation selected in a disposable process. A generator check confirms mixed ASCII/Unicode columns produce equal byte widths. No database regeneration is included.

Verification

• Ruby 4.0.6 via rbenv: eight existing core tests / 22 assertions pass, before and after.
• Existing optional mime-types parity test: nine tests / one pre-existing error on both baseline and patch (newer reference-data extensions absent from the bundled snapshot). No data refresh or test changes.
• Full RuboCop: one pre-existing redundant-disable offense on main and each patch, no additions. Ruby syntax and git diff --check pass.
• No repository tests added/modified under the commissioning repository’s explicit policy; focused checks run externally. Other Ruby versions/platforms were not run locally.

Compatibility

No intended breaking change. ASCII database bytes and lookup results are unchanged; Unicode custom databases must use fixed byte widths, as required by positional reads. The fallback check is simulated on macOS, not a Windows CI claim.

[SamSaffron]

sorry addressed in my big pr