Release 1.1.0

[ipjohnson]

Version bump, changelog, analyzer release tracking, and three documentation
corrections that should not ship as they stand.

Why 1.1.0 and not 1.0.1

Nothing in the public API breaks — but two members are added to the Runtime
package, and adding public API is a MINOR event under semver, not a patch:

• InterceptAttribute.Realm
• DecoratorHelper.Decorate<TService>(…, Type? implementationType) — a new
  overload; the three-argument one is untouched, so generated code from 1.0.0
  still binds.

Three diagnostics also arrive, and 1.0.0 already set the rule for that: "New
ones may be added in a minor release."

DependencyModules.SourceGenerator.Impl does change incompatibly —
LocationModel moved namespace and ModuleEntryPointModel's constructor gained
a required third parameter — but 1.0.0 put those extension points outside the
versioning promise explicitly, and the changelog records what moved.

Upgrade note, stated prominently in the changelog

DM0017 and DM0019 are errors on code that compiled before. Both fire only
on shapes that built green and registered nothing, so no working program breaks —
but a build carrying one goes red on upgrade. NoWarn and .editorconfig both
silence them if the work needs staging.

Separately: an interceptor now applies only to the implementation it was declared
on, so a sibling implementation that was being intercepted stops being
intercepted. It was never asked for.

Documentation corrections

1. The .editorconfig claim was wrong. The reference said
dotnet_diagnostic.DM0005.severity = none has no effect. It works. Verified three
ways on SDK 10.0.302 against a real DM0019:

State	DM0019 errors
no .editorconfig	2
dotnet_diagnostic.DM0019.severity = none	0
.editorconfig removed again	2

What .editorconfig genuinely cannot do is raise a severity — a generator's
diagnostics arrive with theirs already fixed, so
dotnet_diagnostic.DM0018.severity = warning does nothing. Both halves are now
stated, and the DM0018 descriptor comment that told readers to promote it that
way is corrected. That comment was mine, from #40.

2. DM0010/DM0011 were described as never appearing in dotnet build at any
verbosity. They appear at -v detailed.

3. DM0012 described the opposite of what the generator emits. A condition
with nothing to test cannot be false, so the guard is if (true) and the service
registers unconditionally — the page said it never registers.

DM0017, DM0018 and DM0019 now have table rows and sections on the reference,
including the DM0018 note that it cannot be promoted.

Verification

• dotnet build DependencyModules.sln -c Release — 0 warnings
• dotnet test DependencyModules.sln -c Release — 1,830 passed, 0 failed
• scripts/verify-packages.sh 1.1.0 — all checks green for net8.0 and net10.0,
  including packing, consuming from a real feed, and running the consumer app
• vitepress build — clean, with ignoreDeadLinks: false

And the strongest gate available: the four applications written against 1.0.0
during the evaluation — a CLI, a worker daemon, an Avalonia desktop app and a
Native AOT tool, 11 assemblies, 224 tests — were re-run against this build with a
locally packed feed. 218 pass unchanged. All six failures are the tests that were
deliberately pinning the defects this release fixes:

Failing test	Fix that flipped it
AttributeSpellingTests ×3	qualified/aliased attribute lifetime
KnownDefect_AnInterceptorWrapperIsAppliedToEveryRegistrationOfTheInterface	interceptor scoping
KnownDefect_AnOnlyRealmModuleStillAppliesTheOtherModulesInterceptors	realm filtering
ComposingThroughTheAttributeSilentlyResetsNonNullableParameters	module parameter defaults

Zero genuine regressions, and neither new error diagnostic false-positived across
those 11 assemblies.

Publishing

Merging this does not publish. The release workflow triggers on a v* tag, which
is the source of truth for the published version — so git tag v1.1.0 && git push origin v1.1.0 once this is on main.

🤖 Generated with Claude Code

https://claude.ai/code/session_01R9DWh8FWTib6hRYURbypWU