Make HTTP/2 channel creation timeout configurable

[kt2205]

Summary

• The HTTP/2 parent channel creation timeout was hardcoded at 10,000ms with a TODO comment to expose it via cfg2
• When a downstream host becomes unreachable, all concurrent requests to that host queue behind the parent channel bootstrap and block for the full 10s timeout before failing — destroying p95 latency for callers
• This change threads a configurable channelCreationTimeoutMs through: HttpClientFactory.Builder → ChannelPoolManagerFactoryImpl → Http2ChannelPoolFactory → Http2ChannelLifecycle

Usage

new HttpClientFactory.Builder()
    .setHttp2ChannelCreationTimeout(3000)
    .build();

Backwards Compatibility

• Default remains 10,000ms — no behavioral change for existing users
• All existing constructors preserved as @deprecated overloads
• Zero breaking changes

Motivation

In production, a single flaky host that fails SSL handshakes intermittently causes the 10s hardcoded timeout to block all concurrent HTTP/2 requests to that host (due to the parent channel bootstrapping pattern in Http2ChannelLifecycle). A lower configurable timeout (e.g. 3s) allows D2 degrader to mark unhealthy hosts faster, significantly reducing p95 latency impact.

Files Changed

File	Change
Http2ChannelLifecycle.java	Accept channelCreationTimeoutMs as constructor param
Http2ChannelPoolFactory.java	Thread timeout through; deprecate old constructors
ChannelPoolManagerFactoryImpl.java	Thread timeout through; deprecate old constructor
HttpClientFactory.java	Add setHttp2ChannelCreationTimeout() builder method

[TylerHorth]

LGTM

[kt2205]

CI Failure Analysis

The Java 11 CI failure is not caused by this PR. Both failing test suites are pre-existing flaky tests on the repo:

test.r2.integ.clientserver.TestAlpnUpgradePromise > testClearTextAndAlpn

• The exact same failure (49+ testClearTextAndAlpn repetitions, setup FAILED) appears in the master branch CI run from May 14 (run 25877139614) — before this PR existed.
• Root cause: ALPN/TLS environment issue on GitHub Actions runners, unrelated to HTTP/2 channel creation timeout changes.

com.linkedin.d2.balancer.clients.RetryClientTest

• This test is in the d2 module. This PR touches zero files in d2 — only 4 files in r2-netty/ are modified.
• Also present in pre-existing master CI failures.

Files changed in this PR

r2-netty/src/main/java/com/linkedin/r2/netty/client/http2/Http2ChannelLifecycle.java
r2-netty/src/main/java/com/linkedin/r2/netty/client/http2/Http2ChannelPoolFactory.java
r2-netty/src/main/java/com/linkedin/r2/transport/http/client/HttpClientFactory.java
r2-netty/src/main/java/com/linkedin/r2/transport/http/client/common/ChannelPoolManagerFactoryImpl.java

Could a maintainer re-run CI or merge given the failures are pre-existing and unrelated to this change? The PR already has an approval from @TylerHorth.