Fix Go client data races

[chlowell]

The client's running bool field is accessed concurrently by Stop() and readLoop() methods without synchronization, creating a data race. This can be prevented by simply changing the field's type to atomic.Bool. I also enabled the go test race detector in the test script so future CI runs will fail when the test suite triggers a race.

[Copilot]

Pull request overview

This PR fixes a data race in the Go JSON-RPC client by converting the running field from a plain bool to atomic.Bool. The race occurred because the field was accessed concurrently by the Stop() method and the background readLoop() goroutine without synchronization. The PR also enables the Go race detector in the test script to catch such issues in future CI runs.

Changes:

• Changed running field from bool to atomic.Bool and updated all access points to use atomic operations (Store(), Load())
• Enabled the -race flag in go test command to detect data races in CI

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
go/test.sh	Added -race flag to enable Go's race detector during test execution
go/internal/jsonrpc2/jsonrpc2.go	Fixed data race by converting running field to atomic.Bool and using atomic operations for all accesses

[SteveSandersonMS]

Thanks for contributing this, @chlowell!

I notice that CI is failing with this change due to it detecting another data race:

==================
    testing.go:1490: race detected during execution of test
=== RUN   TestRpc/should_call_RPC.Models.List_with_typed_result
    rpc_test.go:64: Not authenticated - skipping models.list test
==================
WARNING: DATA RACE
Write at 0x00c000176228 by goroutine 325:
  github.com/github/copilot-sdk/go.(*Client).ForceStop()
      /Users/runner/work/copilot-sdk/copilot-sdk/go/client.go:386 +0x110
  github.com/github/copilot-sdk/go/internal/e2e.TestRpc.func2.1()
      /Users/runner/work/copilot-sdk/copilot-sdk/go/internal/e2e/rpc_test.go:52 +0x2c
  testing.(*common).Cleanup.func1()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/testing/testing.go:1211 +0x140
  testing.(*common).runCleanup()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/testing/testing.go:1445 +0x1e0
  testing.tRunner.func2()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/testing/testing.go:1786 +0x48
  runtime.Goexit()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/runtime/panic.go:636 +0x5c
  testing.(*common).Skip()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/testing/testing.go:1132 +0x78
  github.com/github/copilot-sdk/go/internal/e2e.TestRpc.func2()
      /Users/runner/work/copilot-sdk/copilot-sdk/go/internal/e2e/rpc_test.go:64 +0x2dc
  testing.tRunner()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/testing/testing.go:1792 +0x180
  testing.(*T).Run.gowrap1()
      /Users/runner/hostedtoolcache/go/1.24.13/arm64/src/testing/testing.go:1851 +0x40

Previous read at 0x00c000176228 by goroutine 327:
  github.com/github/copilot-sdk/go.(*Client).startCLIServer.func1()
      /Users/runner/work/copilot-sdk/copilot-sdk/go/client.go:1102 +0x34

from https://github.com/github/copilot-sdk/actions/runs/22418728180/job/65004395091?pr=586

If we're adding this detection, could you also update other places that would trigger it? Thanks!

[chlowell]

Sorry I missed that one; I haven't been able to repro it on my machine with the current test suite. I added a test that reliably triggers it and other races in copilot.Client, which I've now fixed. I also fixed some process leaks that came up while I was working on synchronization. The new synchronization is a bit involved because some data shared by Start and [Force]Stop is already synchronized and a simple mutex covering all three methods won't work because ForceStop needs to function concurrently with Start and Stop. I was aiming for a smallish fix; let me know what you think.