atenet: give resume callers independent park budgets

[Du Bin (dubin555)]

Fixes #613

Summary

• replace the fixed-budget singleflight with an explicit per-Actor shared flight
• give every caller its own parking deadline while extending shared execution for late joiners
• wake an accumulated long retry delay when fresh demand arrives without starting a retry loop per request
• document per-request budgets, retry behavior, and expired-flight handoff

Testing

• Tests pass

• Appropriate changes to documentation are included in the PR

• go test ./cmd/atenet/internal/router/... -count=1

• go test ./cmd/atenet/internal/router/ingress -run LateJoiner -count=100

• go test -race ./cmd/atenet/internal/router/ingress -count=1

• go vet ./cmd/atenet/internal/router/...

• hack/verify/golangci-lint.sh

• git diff --check

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[Du Bin (dubin555)]

Aditya Shantanu (@aditya-shantanu) I implemented the per-caller budget behavior described in #613 while preserving a shared per-Actor resume flight. Late joiners now keep their own wait budget and extend the shared flight deadline; the regression tests cover a late caller outliving the original leader and avoiding an accumulated long backoff. Would you be willing to review this approach?