fix: align semconv import with OTel SDK schema URL (controller CrashLoop)

[bdchatham]

TL;DR

Controller pods crash on startup after the #100 image bump. Root cause is a semconv schema-URL mismatch from the OTel work in #97/#98, not #99. One-character import bump fixes it; verified locally.

Crash

{"level":"error","ts":"...","logger":"setup",
 "msg":"Failed to initialize OTel MeterProvider",
 "error":"building OTel resource: conflicting Schema URL:
         https://opentelemetry.io/schemas/1.40.0 and
         https://opentelemetry.io/schemas/1.26.0",
 "stacktrace":"main.main\n\t/workspace/cmd/main.go:107\n..."}

Root cause

resource.Merge(resource.Default(), resource.NewWithAttributes(semconv.SchemaURL, ...)) rejects merging resources with differing non-empty schema URLs.

• resource.Default() reports schema URL https://opentelemetry.io/schemas/1.40.0 — the schema embedded in the OTel SDK (we're on go.opentelemetry.io/otel/sdk v1.43.0).
• cmd/telemetry.go:14 hardcoded semconv "go.opentelemetry.io/otel/semconv/v1.26.0" which contributed https://opentelemetry.io/schemas/1.26.0.

Bumping to semconv/v1.40.0 aligns the two schema URLs. All three symbols used (semconv.SchemaURL, ServiceName, ServiceVersion) are stable across semconv versions — drop-in.

Diff

-	semconv "go.opentelemetry.io/otel/semconv/v1.26.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.40.0"

No go.mod / go.sum churn — semconv subpackages are siblings under the otel module, already resolved.

Verified locally

• go mod tidy — no changes.
• go build ./... — clean.
• make lint — 0 issues.
• make test — green.

Current prod impact

• Deployment has 4 pods instead of 3 (1 CrashLooping new pod + 3 healthy old pods on pre-bump image). Rolling update stuck at maxUnavailable=0.
• Existing chains (pacific-1, atlantic-2, arctic-1) are unaffected — old pods continue reconciling.
• Autobake M2b is blocked — old pods don't populate .status.internalService.

Apply plan

1. Merge this PR.
2. ecr.yml workflow auto-publishes the new image on push to main.
3. Open a chore/bump PR bumping config/manager/manager.yaml to the new SHA. (I'm happy to do that one-liner once this lands.)
4. Flux reconciles → new pods roll out cleanly → old pods terminate.

🤖 Generated with Claude Code