[dajiaohuang]

Problem

Diagnostics currently use UTF-8 byte counts for LSP character positions. With the default UTF-16 position encoding, this puts underlines on the wrong characters after non-ASCII comments and can extend ranges past the end of a line.

For example, enable keyword_forbidden_wire_reg in .svlint.toml, initialize svls with the fixture directory as rootUri and working directory, then open:

module test;
  /* 中文 */ wire a;
endmodule

The stdio textDocument/publishDiagnostics response reports line 1, characters 15..19 for wire; the correct UTF-16 range is 11..15. A comment containing a non-BMP character has the same byte/code-unit mismatch.

Change

Convert parser/linter byte offsets to UTF-16 columns. Convert both endpoints rather than adding byte lengths to a UTF-16 start column, including the parser-error range ending at the end of a line. Add regression tests for BMP/non-BMP comments, line transitions and Unicode range endpoints.

Validation

• cargo install --path . --locked --root <isolated-install>
• cargo test --locked and cargo test --release --locked: 2 tests passed
• Own isolated stdio client: initialize/initialized, didOpen/full didChange/didClose, shutdown/exit; ASCII/BMP/non-BMP diagnostics, Unicode text-rule spans, parser errors, disabled linter, empty/default configuration, excluded paths, include paths and defines (11 scenarios)
• The Chinese-comment stdio range assertion fails on the base and passes with this change; the non-BMP case and both Unicode endpoint scenarios also pass
• git diff --check

Tested on Windows with rustc 1.98.1. Linux/macOS, GUI editor integrations and dynamic plugins were not tested.