feat(session): add single-connection transport

[cardyok]

Summary

• add the v2 bidirectional gRPC session transport while keeping v1 as the default
• multiplex commands, results, and heartbeats with bounded messages, handshake deadlines, and heartbeat cancellation
• preserve session protocol selection across gpud up and auto-update, and reconnect either transport after token rotation
• send the control-plane origin metadata required by the gateway CSRF filter

Rollout

v1 remains the default. v2 is enabled explicitly with --session-protocol=v2 only after the matching Lepton gateway and gpud-manager changes are deployed. auto falls back only when the v2 service returns gRPC UNIMPLEMENTED.

Related Lepton MR: https://gitlab-master.nvidia.com/dgxc-lepton/lepton/-/merge_requests/7763

Verification

• go build ./...
• go test -race -gcflags='all=-N -l' ./pkg/session/... ./pkg/gpud-manager/systemd ./pkg/config ./cmd/gpud/command ./cmd/gpud/run ./cmd/gpud/up -count=1
• custom golangci-lint on all changed packages: 0 issues
• normalized Lepton/GPUd protobuf schemas and deterministic wire golden tests match

Full go test -gcflags='all=-N -l' ./... -count=1 is blocked by existing environment-sensitive tests: Docker ping does not time out when no daemon is available, and the pstore restricted-file assertion is not valid under the current runner permissions. All changed-package tests pass.

DEV end-to-end rollout validation is deferred until the paired Lepton changes are available.

Jira

LEP-5352

[codecov]

Codecov Report

❌ Patch coverage is 91.66667% with 30 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.72%. Comparing base (f0e21ed) to head (3101798).

Files with missing lines	Patch %	Lines
pkg/session/session_v2.go	89.92%	20 Missing and 7 partials ⚠️
pkg/gpud-manager/systemd/systemd.go	91.30%	2 Missing ⚠️
pkg/session/update.go	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1270      +/-   ##
==========================================
+ Coverage   90.68%   90.72%   +0.03%     
==========================================
  Files         351      353       +2     
  Lines       26997    27346     +349     
==========================================
+ Hits        24482    24809     +327     
- Misses       1702     1719      +17     
- Partials      813      818       +5     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.