Closes #1112

What

Add coverage-only NETWORK_COVERAGE_TIMEOUT_MULTIPLIER to tests/support/hermetic_transport_fixture.h so Http2ClientHermeticTransportTest (and other hermetic-fixture users) complete the SETTINGS handshake under -fprofile-arcs -ftest-coverage. Round-5 structural fix per PR #1111 diagnosis.

Change Type

• Test infrastructure (fixture only; no src/ changes; no new TEST_F)

Affected Components

• tests/support/hermetic_transport_fixture.h — define multiplier (default 1); apply to wait_for default-argument timeout.
• tests/support/hermetic_transport_fixture.cpp — scale make_loopback_tcp_pair accept-handler deadline.
• tests/support/CMakeLists.txt — define NETWORK_COVERAGE_TIMEOUT_MULTIPLIER=5 PUBLIC on network_test_support when ENABLE_COVERAGE is ON.

Why

PR #1111 diagnosis (docs/coverage/http2_client_round-4-diagnosis.md) confirmed all 7 Round-3 dispatcher TEST_F (and ~30 other fixture-based tests) FAILED at the wait_for(... settings_exchanged() ..., 3s) gate: 5.1-5.4s observed under coverage vs 100-300ms in Debug/Release. Three consecutive rounds (PR #1108 Round 2, PR #1109 Round 3) merged green in Debug/Release CI but produced exactly 0pp delta on src/protocols/http2/http2_client.cpp (still 18.75% line / 9.91% branch, byte-identical LH/LF/BRH/BRF). Without this structural fix, every future TEST_F added to exercise dispatcher branches will repeat the 0pp pattern.

Related Issues

• Closes infra(test): add COVERAGE_TIMEOUT_MULTIPLIER for hermetic_transport_fixture wait_for (Round 5, Part of #1106, #953) #1112
• Part of test(http2): expand http2_client.cpp error-branch coverage using Phase 2 substrate (round 2) #1106 (Round 1/2/3 coverage push on http2_client.cpp)
• Part of Expand unit test coverage from 40% to 80% #953 (umbrella coverage epic)
• Unblocks test(grpc): expand client.cpp error-branch coverage using Phase 2 substrate (round 2) #1107 (grpc client coverage uses the same fixture pattern)
• Diagnosis source: PR docs(coverage): diagnose http2_client 0pp coverage delta after Round 1+2+3 (#1110) #1111 / closed issue test(http2): diagnose 0pp coverage delta after Round 1+2+3 (Part of #1106, #953) #1110

Where

No src/ changes. No new TEST_F.

How

Implementation

1. Header gates the macro behind #ifndef NETWORK_COVERAGE_TIMEOUT_MULTIPLIER ... #define ... 1 ... #endif so non-coverage TUs see the default 1.
2. wait_for default-argument timeout becomes std::chrono::seconds(2) * NETWORK_COVERAGE_TIMEOUT_MULTIPLIER — outside coverage that resolves to 2s; under coverage it resolves to 10s.
3. The cpp's make_loopback_tcp_pair deadline scales the same way.
4. Build system sets NETWORK_COVERAGE_TIMEOUT_MULTIPLIER=5 PUBLIC on the network_test_support static lib only when ENABLE_COVERAGE is ON. PUBLIC propagates through target_link_libraries(... PRIVATE network::test_support) so consuming test executables compile fixture-using TUs with the same macro value.
5. The fixture-wide wait_for(... settings_exchanged() ..., 3s) call sites in tests/unit/*_branch_test.cpp consume the scaled default; explicit-3s call sites are not modified in this PR (per issue scope discipline — the 3s explicit budgets are caller-provided constants that already exceed observed Debug/Release latency, and the issue's primary lever is the default-argument path that the dispatcher TEST_F use).

Testing Done

• Local diff inspection: confirmed every fixture-internal literal-second budget scales with the multiplier.
• Build-system change scoped to network_test_support (test target only); src/ and production targets are not touched.
• Local toolchain (cmake/Ninja) is available but a full local build would require fetching vcpkg dependencies — relying on CI for the Debug/Release matrix verification.

Test Plan for Reviewers

1. Inspect the wait_for default-argument diff and confirm it scales with NETWORK_COVERAGE_TIMEOUT_MULTIPLIER.
2. Verify the build-system change is scoped to the test_support target (not src/ or production).
3. Check Debug + Release matrix CI is green on the PR (multiplier is 1, no behaviour change).

Acceptance Criteria

• Macro NETWORK_COVERAGE_TIMEOUT_MULTIPLIER exists in tests/support/hermetic_transport_fixture.h, defaults to 1, set to 5 only when ENABLE_COVERAGE is ON.
• Every fixture-internal wait_for(...) budget reachable via the default argument scales with the multiplier; make_loopback_tcp_pair accept-handler deadline scales likewise.
• Debug and Release matrix CI on the PR pass (no behaviour change off coverage).
• POST-MERGE — coverage workflow on develop measures >= +20pp line AND >= +10pp branch on src/protocols/http2/http2_client.cpp vs pre-merge sha (baseline: 18.75% line (108/576), 9.91% branch (97/979) from PR test(http2): stabilize coverage and expand dispatcher branches (#1106) #1109 sha d5378644). Cannot be validated in PR-time CI — PR-time matrix is Debug/Release only. Verified by the post-merge coverage workflow on develop.
• POST-MERGE — Round-3 dispatcher TEST_F (Http2ClientHermeticTransportTest.ConnectCompletesSettingsExchangeWithMockPeer plus the 7 dispatcher cases) PASS under instrumentation. Verified post-merge by the coverage workflow.
• No new TEST_F introduced in this PR.

Breaking Changes

None. Multiplier defaults to 1 outside coverage builds; Debug/Release behaviour is unchanged.

Rollback

git revert is sufficient. Test infrastructure only.

Risks

• The 7 dispatcher TEST_F may add 50+ s combined to the coverage workflow runtime once they actually complete (vs aborting at 5s today). Diagnosis explicitly accepts this trade-off.
• If post-merge coverage delta does not clear the >= +20pp / >= +10pp gate, do NOT close issue infra(test): add COVERAGE_TIMEOUT_MULTIPLIER for hermetic_transport_fixture wait_for (Round 5, Part of #1106, #953) #1112; reopen the diagnosis to investigate whether more call sites need the multiplier or whether Option C (friend-injected process_frame) is needed instead.