Add net10 IPAddress.IsValid and IsValidUtf8

[SimonCropp]

Two members, both net10 — and the shortlist entry named only IsValidUtf8, but IsValid(ReadOnlySpan<char>) arrived alongside it and would be odd to ship without.

A third net10 member cannot be added

Parse(ReadOnlySpan<byte>) is also new in net10, and is the obvious companion to the TryParse(ReadOnlySpan<byte>) Polyfill already has. It cannot be polyfilled: GuidPolyfill already declares Parse(ReadOnlySpan<byte>), and because static extension members carry no receiver in their emitted signature, the two are CS0111 Type 'Polyfill' already defines a member called 'Parse' with the same parameter types. Differing return types do not help.

This is the third time this constraint has bitten — after the three Math.BigMul overloads in #594 — so it is //Note:d on TryParse(ReadOnlySpan<byte>), the nearest member that did ship, naming the specific clash.

Worth recording how it surfaced: dotnet build of Polyfill.csproj passed, because that project does not define FeatureMemory and so compiles IPAddressPolyfill.cs away entirely. Only the Consume build across all 22 TFMs caught it. Consume is the real gate for this repo, not the library project.

Verified against net11 rather than assumed equivalent

A 31-input corpus, chosen for the places these parsers historically disagree — partial IPv4 (1.2.3, 1.2, 1), leading and trailing whitespace, 01.02.03.04, IPv6 with scope ids (fe80::1%eth0, fe80::1%12), bracketed ([::1]) and port-suffixed (1.2.3.4:80) forms, ::ffff:1.2.3.4, and full-width (１.２.３.４) and Arabic-Indic (١.٢.٣.٤) digits:

• IsValid(span) agrees with TryParse(string, out _) on every input — 0 mismatches.
• IsValidUtf8(bytes) agrees with IsValid(chars) on every input — 0 mismatches.
• Malformed UTF-8 (FF, a truncated C3, an overlong C3 28, a surrogate encoding ED A0 80) is rejected, and Encoding.UTF8.GetString + IsValid agrees, since the replacement characters do not parse.
• An empty span is not valid, for either overload.

So the polyfill is TryParse(span.ToString(), out _) and the UTF-8 equivalent, and the tests use TryParse(string) as the oracle — an API that exists unchanged on every framework Polyfill targets, and which the BCL's own IsValid matches.

Allocation note, applied to the whole file

Both new members go through the string overload, which allocates where the BCL validates straight from the span. That is //Note:d — and the same note is added to the three pre-existing members in this file (Parse(ReadOnlySpan<char>), both TryParse span overloads), which have exactly the same characteristic and were previously undocumented. Documenting only the new ones would have left the section inconsistent.

Verification

Solution clean in Release, Consume clean across all 22 TFMs, tests green on net11.0 (1730), net10.0 (1730), net9.0 (1730), net8.0 (1727), net462 (1676), plus PublicTests, EmbeddedTests, UnsafeTests, NoRefsTests and NoExtrasTests. net9.0 is inside the polyfill's window, so it and net10.0 run the identical corpus against the two implementations.

API count 1151 → 1153.