Expose disposeInactivePoolsInBackground for HttpClient connection pool

[seonwooj0810]

Fixes #4165

Problem

HttpClientProperties.Pool exposes several Reactor Netty connection-pool knobs (max-idle-time, max-life-time, eviction-interval, …) but not ConnectionProvider.Builder#disposeInactivePoolsInBackground(disposeInterval, poolInactivity).

The existing eviction-interval (evictInBackground) only evicts idle connections within a pool; it does not remove the pool entries themselves from Reactor Netty's internal channelPools map. When downstream instances churn — e.g. Kubernetes pods rolling, so the remote IP:port (and therefore the PoolKey) keeps changing — entries accumulate and are never reclaimed, which manifests as a slow memory leak (see #4165 for the analysis and a linked Reactor Netty report).

Change

Expose the missing knob, mirroring how the other pool properties are wired:

• Add inactive-pool-dispose-interval and pool-inactivity (both Duration, null/disabled by default) to HttpClientProperties.Pool.
• In HttpClientFactory#buildConnectionProvider, call builder.disposeInactivePoolsInBackground(...) only when both are set, so default behavior is unchanged.

spring.cloud.gateway.server.webflux.httpclient.pool.inactive-pool-dispose-interval=5m
spring.cloud.gateway.server.webflux.httpclient.pool.pool-inactivity=5m

Test evidence

Extended the existing GatewayAutoConfigurationTests#nettyHttpClientConfigured binding test to set and assert both new properties.

./mvnw -pl spring-cloud-gateway-server-webflux -Dtest='GatewayAutoConfigurationTests#nettyHttpClientConfigured' test
Tests run: 1, Failures: 0, Errors: 0, Skipped: 0
BUILD SUCCESS

checkstyle + spring-javaformat validate pass. Commit is DCO signed-off.

Verification done

1. No in-flight PR (searched open/closed PRs for disposeInactive / 4165 — none). 2. No self-claim comments on the issue. 3. Code-focused — touches only .java. 4. Confirmed the property/builder call is absent on current main. 6. Triage gate N/A (spring-cloud/, not spring-projects/).