fix: harden logtail push lifecycle

[XuPeng-SH]

What type of PR is this?

• BUG
• Improvement
• Test and CI

Which issue(s) this PR fixes:

Fixes #25644

What this PR does / why we need it:

Hardens the TN logtail push lifecycle under normal unhappy-path conditions.

• Makes subscription completion generation-aware so canceled or superseded phase results cannot publish stale responses.
• Honors caller cancellation in the logtail client request queue and in disttae readiness waiting.
• Isolates a congested session from global incremental publication.
• Releases queued logtail batches during server shutdown and establishes a bootstrap waterline before the first incremental event.
• Detects and removes idle disconnected morpc sessions with a low-frequency probe.
• Adds deterministic in-memory, real morpc integration, concurrent lifecycle, shutdown, ownership, bootstrap, boundary, and race coverage.

Validation

• go test -race ./pkg/vm/engine/tae/logtail/service ./pkg/vm/engine/tae/logtail ./pkg/vm/engine/disttae
• Repeated race coverage for lifecycle and generation tests.
• git diff --check

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[XuPeng-SH]

Re-reviewed the latest head be0fc6d627. The only PR-local delta after my last pass is the mock logtail session lifecycle fix in pkg/vm/engine/test/testutil: testClientSession now has a real session context rooted in the transport context, Close() cancels it, and Write() terminates on either the caller context or the session/transport shutdown. That closes the remaining test-fixture gap where a mock session could keep looking alive and keep a send path blocked after transport teardown, so the unhappy-path tests now model the production morpc/session lifetime much more faithfully.

I reran the focused packages on the latest head (./pkg/vm/engine/test/testutil, ./pkg/vm/engine/tae/logtail/service, ./pkg/vm/engine/disttae) plus race for ./pkg/vm/engine/test/testutil and ./pkg/vm/engine/tae/logtail/service, and I do not see a new high-confidence correctness or Q1/Q2/Q3 blocker on this head. GitHub still will not let me APPROVE my own pull request, so I am leaving this as another approve-level review comment.

[gouhongshen]

Review result: REQUEST_CHANGES. I reviewed the lifecycle, ordering, cancellation, callback ownership, slow-session isolation, heartbeat/bootstrap protocol, and regression tests. I found one P1 blocker below. The issue is a protocol-ordering violation under a congested session, so this cannot be approved as-is. Validation: git diff --check passed; the focused logstore/sm unit and race tests passed. The service/disttae package tests could not compile in this checkout because the environment lacks required cgo headers (usearch.h/roaring.h/xxhash.h).