chore: fix panic

[thelostone-mc]

Problem

offset_to_line in src/check/utils.rs converts a finding's source location to a line number for display. It iterated content.chars().enumerate() — character indices — but compared them against start, which comes from solang Locs and is a byte offset. The two only coincide for pure-ASCII files, which is why this went unnoticed.

For files containing multi-byte UTF-8 (emoji, non-Latin comments), there are two symptoms:

1. Panic: if a finding's byte offset exceeds the file's total character count (any finding near the end of a file with enough multi-byte content), the loop runs out of characters and hits unreachable!("content.len() > start"), crashing the entire scopelint check run:

   thread 'main' panicked at src/check/utils.rs:212:5:
   internal error: entered unreachable code: content.len() > start
   

   Repro: a file with a // 🚀🚀🚀... (300 emoji) comment followed by a naming violation.
2. Wrong line numbers: with less multi-byte content the loop merely returns late, attributing findings to later lines than the truth (e.g. a violation on line 6 reported on line 8).

Fix

Iterate with char_indices() instead, which yields byte offsets, making the comparison byte-to-byte. The debug_assert!/unreachable! are replaced with a defensive fallback returning the last line when start is at or past EOF.

This was the only chars().enumerate() offset comparison in the codebase — other modules (e.g. inline_config.rs) already use char_indices() correctly.