feat(reader): remove ParseNode infrastructure

[baywet]

Summary

• remove the ParseNode wrapper infrastructure from reader parsing
• parse directly with System.Text.Json JsonNode/JsonObject APIs and explicit ParsingContext propagation
• prune unused JsonNode helper methods
• refresh benchmark report artifacts

Benchmark Comparison vs Vincent Biret's Last Update (eb1258a)

Hardware difference:

• Vincent's run: Intel i7-1185G7 3.00GHz (x64), .NET SDK 8.0.418, .NET 8.0.24 runtime
• Current run: Snapdragon X 12-core X1E80100 3.40 GHz (ARM64), .NET SDK 10.0.300, .NET 8.0.27 runtime

Results comparison (Vincent → Current):

Benchmark	Time Change	Memory Change
PetStoreYaml	405.2 μs → 276.3 μs (32% faster)	361.39 KB → 305.91 KB (15% less)
PetStoreJson	173.0 μs → 112.8 μs (35% faster)	223.53 KB → 168.05 KB (25% less)
GHESYaml	951,103.8 μs → 608,668.3 μs (36% faster)	345343.16 KB → 250121.85 KB (28% less)
GHESJson	435,425.1 μs → 244,147.6 μs (44% faster)	206865.71 KB → 107293.42 KB (48% less)

Gen0/Gen1 GC collections: Similar or slightly lower across all benchmarks.

Key takeaway: The refactored code shows significant improvements in both execution time (30-45% faster) and memory allocation (15-50% less). However, note this is not an apples-to-apples comparison due to different hardware (ARM vs x64) and runtime versions. For a true performance assessment, you'd need to rerun Vincent's baseline commit on the same Snapdragon X hardware.

Validation

• dotnet build Microsoft.OpenApi.slnx --no-restore --verbosity minimal
• dotnet test Microsoft.OpenApi.slnx --no-build --verbosity minimal
• dotnet run -c Release (from performance\benchmark)

[sonarqubecloud]

Quality Gate failed

Failed conditions
79.8% Coverage on New Code (required ≥ 80%)
43.5% Duplication on New Code (required ≤ 3%)
B Reliability Rating on New Code (required ≥ A)

See analysis details on SonarQube Cloud

Catch issues before they fail your Quality Gate with our IDE extension SonarQube for IDE