- Reading once the file located in `cwd` in Runtime::run
- Read only the file content
- Diff subset of this file content in each worker.
- Basic test integration

Found a case not correctly handle, when the project doesn’t have a
file. This case will cover when we are handling the cli args. 

The diff operation is done outside of rayon, but it can be done in each
worker. Until we have a complete integration suite, it won’t
be refactored. I wanted a working version, not a perfect design.

Next commit will be showing errors when any of the issues are found. Currently is being printed as simple logs in the stdout.

Now we can test correctly the following situations:
- No file but not cli provided: No suppression required.
- No file but cli suppress_all provided: Suppression init and creation
- File but not provided any args: filtering working as expected

Pending to not show this error if an cli arguments is passed.

The new hash map is only saved in the following situations:
- If file doesn’t exist , `supress-all` is provided, and a diff is detected, then the file is created.
- If file exist,  `supress-all` or `prune-all` is provided, and a diff is detected, then the file is updated.

The rest of situations, the linter doesn’t make the effort to save the new data. Saving the new suppression data hide the reporting of all diff detected.

Integration test are spplited on two kind:
- Snapshot testing that only check the output is the expect in some situations
- Serialization and desearilazation tests

As the second one required handle files, thoses doesn’t assert the cli output only the file content

…for reading

Removed the need of read twice the file due moving value in Rayon, readings are made in a papaya hash map for concurrent readings.

Suppression file tracking has his own struct instead of create the data through SupressionManager. This  avoid move the value and the need to read again the file. On top of that, this data only make create  in the context of a worker.

The path is clean to only emit the diff to update the current file.

All the heavy work has been moved to the workers. Instead of building from zero the same Hashmap we are doing the following:

- A Read-Only concurrent hash map. From here the workers will read the data to calculate the diff required to update the OG.
- Only in update mode this diff are send through a channel. Doing this as once the file exist we don’t expect a lot of changes between runs.
- Only update the file if the args are present AND diffs are found. Otherwise, we do nothing.

- All JSPlugins errors are being recorded in the file. Left Typescript as it required more changes.
- Add JSPlugin test
- Manual tested the binary in Posthog and oxlint-react-compiler-rules
- Improved errors messages, and added a label to suggest the user what command run to update the file.
- Some basic message in the cli output if the file is created or updated.

- Remove SuppressionId, this information already exist. Probably the reason of the performance regression.
- SuppressionManager inited `apps/oxlint`. This replace the logic if the file has been created or updated. Pending the state if a diff haven’t been found, this will avoid render the file has been updated when it isn’t.
- Integration test for malformed json. The error don’t panic, the files are linted but a message is rendered with the error.
-  `OxcDiagnositc` is transformed to `RuleName`. Pending to use TryFrom instead of From, as it can fail.
- Removed misleading TODO comment

- `—fix —suppress-all`  test without file not generate a `oxlint-suppressions.json` file and fixes are apply.
- `—fix —-suppress-all`test having a initial suppression file, is not behaving correctly as the fix aren’t apply. Are filtered due the suppresion file. Pending to be fixed.
- Using Option instead on initiating a default SuppressionFile struct. This solve an edge case of emitting errors when the file was malformed.

Pending add more test with a combination of fix commands, a test that prune if only `fix` is provided.

…ify file update if the file has been really updated

- Only mentioine the file has been update once the file is write instead of detecting if it exists.
- Filtering the messages with the suppression only after of this fix has been apply, not before.
- Pending more integration tests handling different of fix

Adds LSP support to show `oxlint-suppressions.json`-suppressed violations
as faded/de-emphasised diagnostics (`DiagnosticTag::UNNECESSARY`) rather
than hiding them entirely, with a toggle and configurable severity.

New LSP options (sent via `initializationOptions`):
- `showSuppressedViolations: bool` (default: `false`)
- `suppressedViolationSeverity: "original" | "hint" | "warning" | "error"`

Infrastructure changes:
- `Message.suppressed: bool` field added (default `false`)
- `SuppressionManager::mark_suppressed()` marks violations in place
  instead of dropping them (LSP counterpart to `suppress_lint_diagnostics`)
- `LintRunner::run_source_marking_suppressed()` for the LSP mark path
- `SuppressionTracking` uses `Arc<HashMap>` instead of `papaya` for
  simpler sharing semantics

CLI behaviour is unchanged (suppressed violations are still dropped).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>