chore: Modify powercfg output parse logics

[filzrev]

This PR intended to fix BenchmarkRunnerAcquiresWakeLock tests are failed with timeout error when running on non-english Windows environment.

powercfg /requests command output is localized based on UI language setting of Windows.
So it need to handle localized message.

So I've added IsNoneToken helper method to check None. equivalent string.

---

The added language mapping are generated by LLM.
And it's confirmed powercfg /requests output message data source per language.

There are some languages that can't find reliable message sources.
So test still failed on these language environment.

[timcassell]

Good diagnosis, but the language table isn't needed — that line is positionally determined. After a RequestType token the next line can only be a [...] requester or the "none" marker; a Reason only ever follows a RequesterName. ParseRequestType already relies on this (it unconditionally Takes None there), so matching the word is a redundant extra constraint that happens to be the locale-dependent one.

Drop IsNoneToken and track the previous token in Tokens() instead:

else if (line[0] == '[')
{
    int pos = line.IndexOf(']');
    yield return new Token(TokenType.RequesterType, line.Substring(1, pos - 1));
    previous = TokenType.RequesterName;
    yield return new Token(TokenType.RequesterName, line.Substring(pos + 2));
}
else if (previous == TokenType.RequestType)
{
    // Any single line directly after a request type header is the localized "None."
    previous = TokenType.None;
    yield return new Token(TokenType.None, line);
}

I ran both versions over the same input:

English:              PR #3238  OK    positional  OK
German (in table):    PR #3238  OK    positional  OK
Hungarian (absent):   PR #3238  FAIL  positional  OK
Czech (absent):       PR #3238  FAIL  positional  OK

Identical where this PR works, plus every locale it doesn't — so the caveat in the description goes away. It's also a smaller diff, and avoids a table that CI can't verify (English-only) and that has real ambiguity in it (Ninguna. vs Ninguno.).

Two adjacent things, either PR or follow-up:

• pong.Set() in WakeLockTests.WaitForBenchmarkRunningAndGetPowerRequests should be in a finally. A parse failure currently costs a full testTimeout and reports TimeoutException instead of the real InvalidCastException — that's why this presented as a timeout.
• PowerRequestsParser has no unit tests. A couple of canned-string cases would run everywhere, whereas WakeLockTests needs Windows and elevation.

Reviewed by Claude (Opus 5), posted by @timcassell.