#nowarn in the editor does not ignore warning if --warnaserrors flag is set, but compiler does

(edited, it is not related to `--warnon`, but to `--warnaserrors`)

If you have `--warnaserror:xxxx`, and on file-level you disable that same warning, you still see the errors in that file, while the compiler succeeds. This is confusing.

## Repro steps (1)

1. Add "25" to the _Warnings as errors_:
2. Add `#nowarn "25"` to a source file.
2. Write code that triggers this warning:

## Repro steps (2)

1. Add `warnon:1182` in project settings:
2. Add `FS1182` to the "warnings as errors":
3. Inside a source file, add `#nowarn:1182` at the top
4. Have a line with an unused variable


**Expected behavior**

Inside a file, the editor should consider the `#nowarn` directives.

**Actual behavior**

Apparently, the editor highlighting tooling is aware of the `--warnaserror` compiler option, but not of the `#nowarn` directive in the file, and shows the error, while `fsc` doesn't show an error.

* It shouldn't show the red squiggles
* It should (probably) still show the "unused" warning in the mouseover
* It should keep the ident greyed out

**Known workarounds**

Don't use these options, or fix your errors. But as seen in #9795, that's not always possible.

**Related information**

* It doesn't matter whether you use the UI setting, or add `--warnaserror:XXXX` to the addition fsc options

----

Not really related anymore, but FYI, since 1182 is a "special" warning that does not get triggered on any warning level, unless you add the `--warnon:XXXX` to the options.

In this reaction, back in 2015 (!) (https://github.com/dotnet/fsharp/issues/422#issuecomment-99961730), @dsyme mentions that opt-in for FS1182 was deliberate. It was done before moving to github, so I cannot find the decision/discussion or changes.

> Yes, I do recall this, it's good to have it recorded (and yes, this was one of the reasons /warn:1182 was made opt-in)