chore: format the three drifted test files and check formatting in CI

[dborgards]

Both side findings from #87.

1. dotnet format fails on main

Reproduced before touching anything: import ordering in Nfr006ErrorArchitectureTests.cs, UdsClientTests.cs and UdsTransferTests.cs, plus one whitespace complaint. Nothing enforced formatting, so it drifted — and three agents in a row reported it while working on unrelated issues, each spending effort deciding it was pre-existing and not theirs.

Two fixes are simply the formatter's output. The third is not, and it is worth a look:

-                memorySize: new byte[] { (byte)payload.Length }, // sizeWidth=1 caps at 255, but
-                // the ECU only echoes width from ALFI; we simply need a well-formed 1-byte size
-                // for the wire encoding test (this test does not verify size semantics).
+                // sizeWidth=1 caps at 255, but the ECU only echoes width from ALFI; we simply
+                // need a well-formed 1-byte size for the wire encoding test (this test does not
+                // verify size semantics).
+                memorySize: new byte[] { (byte)payload.Length },

dotnet format wanted the continuation lines aligned under the trailing comment, at column 66 and well past the line width — technically consistent, materially worse to read. Moving the comment above the argument satisfies the formatter and reads better. No .editorconfig rule was relaxed to make anything pass.

2. The stale pragma in ControllableBus.cs

Already fixed in #87, which started raising the event that #pragma warning disable CS0067 // Never raised claimed nobody listened to. Nothing left to do.

The new format job

dotnet format --verify-no-changes now runs in CI as its own job, deliberately not inside build-test. The three OS legs are the required status checks on main; a misplaced blank line has no business blocking a merge. This reports, it does not gate. If that proves too weak, promoting it is a ruleset change rather than a workflow change.

Verified

• dotnet format CanKit.Pro.sln --verify-no-changes → exit 0
• dotnet test CanKit.Pro.sln -c Release → 412 passed, 0 failed
• No production code touched; the API approvals do not move.

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
Changes are limited to CI workflow steps and test-file formatting; no production or API behavior is modified.

Overview
Adds a dedicated format CI job that runs dotnet format --verify-no-changes on Ubuntu. It is intentionally not part of the required build-test matrix so formatting drift is reported without blocking merges on the three OS legs.

Windows build-test no longer runs one solution-wide dotnet test. It runs net10.0 with XPlat coverage and test-results-net10.0.trx, then net48 (Windows only) without --collect and with test-results-net48.trx, avoiding TRX overwrite and the bogus “XPlat Code Coverage” collector error on .NET Framework.

Three test files are brought in line with the formatter: using order in Nfr006ErrorArchitectureTests, UdsClientTests, and UdsTransferTests, plus moving a multi-line comment above the memorySize argument in UdsTransferTests so dotnet format passes without ugly line wrapping.

^{Reviewed by Cursor Bugbot for commit af57229. Bugbot is set up for automated code reviews on this repo. Configure here.}