[MRG] Buffer CONNECT response bytes from proxy until all HTTP headers are received

[redapple]

Fixes #2491

See #2491 (comment) for rationale.
A simple fix, but ideally, we'd want to use a proper HTTP parser.
Something for another PR I believe.

[codecov-io]

Codecov Report

Merging #2495 into master will decrease coverage by -0.02%.
The diff coverage is 60%.

@@            Coverage Diff             @@
##           master    #2495      +/-   ##
==========================================
- Coverage   83.46%   83.44%   -0.02%     
==========================================
  Files         161      161              
  Lines        8780     8784       +4     
  Branches     1288     1289       +1     
==========================================
+ Hits         7328     7330       +2     
- Misses       1204     1205       +1     
- Partials      248      249       +1

Impacted Files	Coverage Δ
scrapy/core/downloader/handlers/http11.py	91.13% <60%> (-0.71%)	❌

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4ca191e...adb180f. Read the comment docs.

[kmike]

This could have a different algorithmic complexity in pypy; it seems using bytearray could help. I'm not sure if this is a real problem though.

[kmike]

LGTM. Relevant document: http://www.freeproxy.ru/ru/free_proxy/faq/draft-luotonen-web-proxy-tunneling-01.txt:

An example of a tunneling request/response in an interleaved
   multicolumn format:

     CLIENT -> SERVER                        SERVER -> CLIENT
     --------------------------------------  -----------------------------------
     CONNECT home.netscape.com:443 HTTP/1.0
     User-agent: Mozilla/4.0
     <<< empty line >>>
                                             HTTP/1.0 200 Connection established
                                             Proxy-agent: Netscape-Proxy/1.1
                                             <<< empty line >>>
                  <<< data tunneling to both directions begins >>>

so this PR does as specified - waits for an empty line.

[kmike]

This also can be O(N^2) if connectBuffer is extended byte-by-byte, but I'm not sure how to fix it, and likely it shouldn't be a problem as the response should be small.