MOB-43620 Start Xvfb with -noreset to stop the 56-iteration Chrome hang

[devership16]

Jira: https://perforce.atlassian.net/browse/MOB-43620

Problem

Selenium/BPT tests with 1 user per engine fail on exactly iteration 56 and every iteration after it with session not created: DevToolsActivePort file doesn't exist. Chrome itself is fine. It hangs because the virtual display (Xvfb) has stopped responding.

Root cause

• VirtualDisplay starts Xvfb through pyvirtualdisplay, which sets stderr=PIPE and reads that pipe only in stop().
• With 1 user, each Chrome quits before the next one starts. Xvfb then has no clients, so it resets after every iteration.
• On each reset, Xvfb runs xkbcomp. On Ubuntu 24.04 (xkb-data 2.41) that writes ~1.2 KB of Could not resolve keysym warnings to the unread pipe.
• After ~55 resets the 64 KiB pipe is full. xkbcomp blocks on write, Xvfb waits for it forever, and every later Chrome hangs at X11 init.
• With ≥2 users per engine, the browsers overlap, so Xvfb rarely resets. That is why only 1-user engines are affected.

This is the "Chrome finally crashed after 55 iterations" from #1983. #1983 fixed the 50-minute hang that came after that crash. This PR fixes the crash itself.

Fix

Start Xvfb with -noreset (Display(size=..., extra_args=["-noreset"])). Per-client X resources are still freed when each client disconnects. This is the same state Xvfb is already in for tests with ≥2 users per engine.

Testing

• New unit test TestVirtualDisplay.test_xvfb_started_with_noreset.
• Real pyvirtualdisplay 3.0 SmartDisplay on ubuntu:24.04:
  • master: X connection Bzt moves #56 hangs (xkbcomp in do_wait).
  • this branch: Xvfb runs with -noreset, 150/150 OK.
• BlazeMeter cloud (current image), 1 user/engine, 5 min:
  • unchanged script, master 83533938: fails at iteration 56.
  • same script holding one X connection open (prevents resets, same effect as -noreset), master 83534208: 123/123 OK.

Each PR must conform to Developer's Guide.

Quick checklist:

• Description of PR explains the context of change
• Unit tests cover the change, no broken tests
• No static analysis warnings (Codacy etc.)
• Documentation update ('available in the unstable snapshot' warning if necessary) (not needed: no user-facing option)
• Changes file inside site/dat/docs/changes directory, one-line note of change inside (not added)

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.24%. Comparing base (3af0054) to head (f3b6a15).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2034   +/-   ##
=======================================
  Coverage   88.24%   88.24%           
=======================================
  Files          75       75           
  Lines       21247    21247           
=======================================
  Hits        18748    18748           
  Misses       2499     2499           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.