Implements the ability to add custom reason/comment text when using the
--add-noqa command. This allows users to distinguish auto-generated noqa
comments from manual ones and add contextual information.

Usage:
  ruff check --add-noqa --add-noqa-reason "TODO: fix later"

The reason is appended after the rule codes in the noqa comment:
  # noqa: F401, F841 TODO: fix later

Changes:
- Add --add-noqa-reason CLI flag that requires --add-noqa
- Thread reason parameter through add_noqa command chain
- Update NoqaEdit struct to include optional reason field
- Modify write() method to append reason before line ending
- Update all test calls to pass None for backward compatibility

Fixes #10070

Sanitize newlines in --add-noqa-reason to prevent syntax errors

If a user passes a reason containing newline characters (e.g., via
bash $'line1\nline2'), it would break Python syntax by creating
actual line breaks in the noqa comment:

  import os  # noqa: F401 line1
  line2  # <- This becomes code, not a comment!

This commit sanitizes the reason text by replacing all \n and \r
characters with spaces, ensuring the noqa comment stays on one line.

Added comprehensive test coverage for:
- Unix newlines (\n)
- Windows newlines (\r\n)
- Multiple consecutive newlines

Validate --add-noqa-reason for newlines instead of sanitizing

Changed approach from silently sanitizing newlines to explicitly
rejecting them with a clear error message. This follows the "fail
fast" principle - it's better to alert users to problematic input
rather than silently modifying it.

The CLI now validates the reason text upfront and aborts with:
  "error: --add-noqa-reason cannot contain newline characters"

This prevents syntax errors from malformed Python comments while
being explicit about what input is accepted.

Changes:
- Add validation in lib.rs that checks for \n and \r characters
- Remove sanitization code from noqa.rs write() method
- Update test to verify normal reason text works correctly
- Document that validation happens at CLI level

Refactor --add-noqa to accept optional reason value directly

Changed from two separate flags (--add-noqa and --add-noqa-reason)
to a single flag with an optional value for better ergonomics.

Before:
  ruff check --add-noqa --add-noqa-reason "TODO: fix"

After:
  ruff check --add-noqa "TODO: fix"
  ruff check --add-noqa              # still works without reason

Implementation:
- Changed `add_noqa: bool` to `add_noqa: Option<String>` in args
- Use clap's default_missing_value to handle flag without value
- Empty string means flag present without value (no reason)
- Non-empty string is the reason text to append
- Removed separate add_noqa_reason field entirely

This provides a cleaner, more intuitive CLI interface while
maintaining full backward compatibility with the original
--add-noqa behavior.

Add comprehensive CLI tests for --add-noqa feature

Created new test file tests/cli/add_noqa.rs with end-to-end tests:

- add_noqa_without_reason: Verify original behavior still works
- add_noqa_with_reason: Test appending custom reason text
- add_noqa_with_short_reason: Test with short marker like "auto"
- add_noqa_no_violations: Verify no changes when no violations
- add_noqa_with_existing_comment: Test merging with existing noqa
- add_noqa_multiple_violations_same_line: Test multiple codes
- add_noqa_with_newline_error: Verify newline validation
- add_noqa_multiple_files: Test bulk operations across files

These tests verify:
1. Files are correctly modified in-place
2. Reason text is properly appended after rule codes
3. Multiple violations on same line are combined
4. Newlines are rejected with clear error message
5. Works correctly with multiple files

Tests use insta snapshots for both command output and file contents.

Simplify tests to only essential cases for --add-noqa reason

Removed excessive integration tests and unit tests. Added only 2 focused
tests to existing lint.rs test suite:

1. add_noqa_with_reason: Tests appending reason text to noqa directives
2. add_noqa_with_newline_in_reason: Validates newline rejection

Also removed unit test from noqa.rs since integration tests are sufficient.

Since --add-noqa is already well-tested with 7 existing tests covering
base functionality, these 2 new tests are enough to verify the reason
parameter feature.

Fix test formatting to match existing test patterns

- Remove leading whitespace from test input strings
- Change snapshot format from @r###" to @r" to match existing tests
- Input strings should not have indentation since dedent is applied

This should fix CI test failures.

Refactor tests to match existing --add-noqa test patterns

Changes to follow the exact pattern of existing tests:
- Use CliTest::new() + write_file() instead of with_file()
- Use .check_command() which includes standard flags
- Use separate .arg() calls instead of .args()
- Pass specific filename instead of current directory

This should fix remaining CI test failures.

Fix CLI argument handling: require equals syntax for optional value

Changes:
- Add require_equals = true to --add-noqa argument definition
- Update tests to use --add-noqa=value syntax instead of --add-noqa value
- This makes the optional value syntax unambiguous

Usage:
  ruff check --add-noqa                 # no reason
  ruff check --add-noqa="TODO: fix"    # with reason

This should fix CI test, clippy, and formatting failures.

Simplify code to use .then() instead of if/else for clippy

Fix formatting: use then_some instead of then for better idiomaticity

Run cargo fmt to fix formatting

Fix test snapshot indentation

Remove leading spaces from inline snapshot expectations in add_noqa tests.
The actual file content doesn't have leading indentation, so the snapshots
shouldn't either.

Fix add_noqa integration tests

- Update add_noqa_with_reason test to trigger both F401 and F841 violations
  by using a function (F841 only applies to local variables, not module-level)
- Fix error message snapshot for newline validation to match actual output format

Fix clippy warning for inline format arg

Nits, clippy, generate all