Share the build settings, enforce warnings everywhere, measure coverage

[Kentarohakase]

Stacked on #4. The build-infrastructure items from the improvement plan's low-priority list.

Shared build settings

Both project files declared TargetFramework, Nullable and ImplicitUsings separately, and the test project inherited neither AnalysisLevel nor EnforceCodeStyleInBuild nor the warning policy — its code was warning-free by omission rather than by rule. Those settings move into Directory.Build.props; each project file keeps only what is genuinely its own.

Warnings are errors everywhere, not only in Release

TreatWarningsAsErrors was set under Condition="'$(Configuration)' == 'Release'". A local dotnet build accepted what CI rejected, so a warning surfaced after the push rather than before it. Measured before changing it — both projects build clean in Debug and Release, and the test project builds clean with the analyzers and the warning policy applied:

dotnet build -c Debug                                        -> 0 warnings
dotnet build AudioQualityEnhancer.Tests.csproj -c Release \
  -p:TreatWarningsAsErrors=true -p:AnalysisLevel=latest \
  -p:EnforceCodeStyleInBuild=true                            -> 0 warnings

AnalysisMode stays off, deliberately

The plan noted that AnalysisMode is unset, so most CA rules sit at suggestion severity and never become warnings — and therefore never become errors under TreatWarningsAsErrors. I measured the blast radius before deciding:

AnalysisMode	Analyzer warnings	Dominated by
default	0	—
Recommended	~300	CA1305, CA1822, CA1859, CA1805
All	~1130	CA1515, CA2007, CA1031, CA1305

That is a cleanup in its own right. Switching it on inside a build file would block every later change until someone worked through it, so it stays at the default and the reasoning is recorded in Directory.Build.props rather than lost.

The same applies to .editorconfig: adding one that merely restates the defaults dotnet format already enforces would be noise, and adding one that changes severities is the same cleanup by another route.

Coverage is measured for the first time

coverlet.collector has been referenced by the test project all along, but ci.yml never collected anything. The test step now collects it and CI keeps the Cobertura report as an artifact.

Current state, measured locally:

Line rate:   49.1 %
Branch rate: 45.1 %

That number is mostly explained by MainViewModel — about 2,100 lines across eleven partials with no tests, for the reason described in #4.

SDK pinned

global.json pins the 10.0 feature band with rollForward: latestFeature, so a future major SDK cannot silently change the build. CI installs 10.0.x, which satisfies it.

No changelog entry

Nothing here changes behaviour for a user of the application, so the changelog is untouched.

Verification

dotnet format .\AudioQualityEnhancer.slnx --verify-no-changes   -> clean
dotnet build  .\AudioQualityEnhancer.slnx -c Debug              -> 0 warnings, 0 errors
dotnet build  .\AudioQualityEnhancer.slnx -c Release            -> 0 warnings, 0 errors
dotnet test   .\AudioQualityEnhancer.slnx -c Release            -> 239 passed, 0 failed

Coverage collection was run locally with the exact command CI now uses, and it produced coverage.cobertura.xml as expected.