Strip quotes from charset before Encoding.GetEncoding

[mrotondo]

Problem

Decoding an MTOM response whose application/xop+xml root part declares a quoted charset — e.g. Content-Type: application/xop+xml; charset="utf-8"; type="text/xml" — throws:

System.ArgumentException: '"utf-8"' is not a supported encoding name.
   at WcfCoreMtomEncoder.MtomPart.GetStringContentForEncoder(MessageEncoder encoder)
   at WcfCoreMtomEncoder.MtomMessageEncoder.ReadMessage(Stream, Int32, String contentType)

A quoted charset value is valid per RFC 7231 §3.1.1.1 (a parameter value may be a token or a quoted-string), and System.Net.Http.Headers.MediaTypeHeaderValue.CharSet returns the value with the surrounding quotes intact (see dotnet/runtime#42079). That value — "utf-8", quotes included — is then passed to Encoding.GetEncoding, which rejects it.

This happens against real servers: IRS MeF's state services (Apache/Axiom) emit charset="utf-8" on the MTOM root part, so the client crashes before it can read the message.

Fix

Trim surrounding quotes from CharSet before calling Encoding.GetEncoding, mirroring what the encoder already does for the type parameter (p.Value.Replace("\"", "")). Two call sites consumed the raw value:

• MtomPart.GetStringContentForEncoder
• MtomMessageEncoder.CreateStream

Both now use CharSet?.Trim('"') guarded by !string.IsNullOrEmpty(...), so an absent charset still falls back to Encoding.Default (no NullReferenceException).

Tests

Adds a WcfCoreMtomEncoder.Tests project with an xUnit theory that decodes a multipart/related MTOM response through MtomMessageEncoder.ReadMessage with the root part's charset set four ways:

charset	before this PR	after
charset="utf-8" (quoted, lower)	ArgumentException	passes
charset="UTF-8" (quoted, upper)	ArgumentException	passes
charset=utf-8 (unquoted)	passes	passes
(absent)	passes	passes

Note: the test project targets net10.0 with System.ServiceModel.Primitives 10.x; adjust the TFM / package version to match your CI's SDK if needed.

[mrotondo]

Please excuse the noise around the Claude-generated tests (though I do think they'd be good to have as part of the project). The critical piece is the changes to MtomMessageEncoder.cs and MtomPart.cs, which I've also manually tested by creating a small script that exercises both code paths and crashes before the change but works correctly after.