Fix flaky UseReaderWithoutDisposing test on high-latency servers

[pmishchenko-ua]

Problem

CI run 33603340491 (PR #24) failed in test-windows → Run SideBySide tests:

Failed SideBySide.QueryTests.UseReaderWithoutDisposing [44 s]
  SingleStoreConnector.SingleStoreException : Connect Timeout expired. All pooled connections are in use.

PR #24 only edits RELEASE.md, so this is unrelated to the PR — it's a pre-existing, latency-sensitive flake. It failed only on the Windows job, which runs against the remote S2MS managed cluster (aws-oregon), while every local-cluster Linux job passed.

Root cause

UseReaderWithoutDisposing starts MaximumPoolSize + 4 (12) threads against a pool of 8 and deliberately leaves each DbDataReader undisposed. Disposing a connection whose reader still has unread rows requires a session reset (network round-trips) before the connection can be reused. Against a high-latency managed server those resets are slow enough that the 4 excess threads can exceed the default 15s pool-wait timeout (ConnectionTimeout) and throw Connect Timeout expired. All pooled connections are in use. (SingleStoreConnection.cs). Loopback-latency local clusters never hit this.

Note the repo already scales timing-sensitive tests via AppConfig.TimeoutDelayFactor, but that only recognizes AppVeyor/Travis/TF_Build/CircleCI — not GitHub Actions — so timeouts aren't scaled on the GitHub Windows runner. Broadening IsCiBuild would have side effects (it also disables ServerFeatures.Timeout tests), so this change is scoped to the one affected test.

Fix

Give the test a generous pool-wait ConnectionTimeout (60s) so it still exercises pool reuse under contention without being sensitive to connection latency.

Verification

Reproduced and verified locally against MariaDB reached through a TCP proxy that injects 50 ms/chunk/direction latency (simulating the remote cluster's RTT). With latency and load held fixed, only ConnectionTimeout was varied:

Scenario	ConnectionTimeout	Result
Reproduction (through latency proxy)	2s	FAIL — Connect Timeout expired. All pooled connections are in use. (44s, identical to CI)
Fix (through latency proxy)	60s	PASS (1m7s)
Sanity (direct DB, no proxy)	60s	PASS (53ms)

The reproduction produced the exact same exception and duration as the failing CI job, isolating the pool-wait timeout as the cause.