If TWISTED_REACTOR is None, reuse any pre-installed reactor

[Gallaecio]

Fixes #5525

[Gallaecio]

The documentation failure can be ignored, I plan to address that separately.

[codecov]

Codecov Report

Merging #5528 (22526ad) into 2.6 (1c1cd5d) will decrease coverage by 0.53%.
The diff coverage is 100.00%.

❗ Current head 22526ad differs from pull request most recent head 6b4c429. Consider uploading reports for the commit 6b4c429 to get more accurate results

@@            Coverage Diff             @@
##              2.6    #5528      +/-   ##
==========================================
- Coverage   88.81%   88.27%   -0.54%     
==========================================
  Files         163      163              
  Lines       10676    10563     -113     
  Branches     1821     1789      -32     
==========================================
- Hits         9482     9325     -157     
- Misses        918      955      +37     
- Partials      276      283       +7     

Impacted Files	Coverage Δ
scrapy/crawler.py	88.46% <100.00%> (-0.37%)	⬇️
scrapy/utils/reactor.py	76.92% <100.00%> (ø)
scrapy/robotstxt.py	73.33% <0.00%> (-24.20%)	⬇️
scrapy/core/downloader/handlers/__init__.py	83.33% <0.00%> (-9.40%)	⬇️
scrapy/squeues.py	92.85% <0.00%> (-7.15%)	⬇️
scrapy/utils/ssl.py	65.85% <0.00%> (-4.88%)	⬇️
scrapy/utils/defer.py	91.83% <0.00%> (-4.13%)	⬇️
scrapy/pqueues.py	95.86% <0.00%> (-3.31%)	⬇️
scrapy/extensions/memdebug.py	45.00% <0.00%> (-2.62%)	⬇️
scrapy/downloadermiddlewares/cookies.py	95.74% <0.00%> (-2.16%)	⬇️
... and 62 more

[Gallaecio]

Hmm… I assumed Windows would be using one of the Windows-specific reactors, but I was wrong, and select seems to be the only crossplatform reactor (making it suitable for the failing test), so I will skip those tests on Windows.