* FIXES an error discovered while testing #14797:
   https://bugs.torproject.org/14797

Pillow, in Debian jessie, is in the `python-imaging` package (currently
version 2.6.1).  However, on PyPI, Pillow is simply `pillow` (currently
version 2.7.0). Both of these are backwards compatible with the
python-imaging package in Debian wheezy (currently version 1.1.6, and
likely to never get updated, because it is being replaced with pillow).
And any of these will currently work with BridgeDB.

 * FIXES #11345 https://bugs.torproject.org/11345

 * FIXES three failing unittests in test_HTTPServer.py:
    - test_render_GET_RTLlang
    - test_render_GET_RTLlang_obfs3
    - test_render_GET_vanilla

This should hopefully make copying the bridge lines elsewhere less
error-prone for users.

 * CHANGES bridge.html template to inline replace '\n' with '<br />' in
   the bridge lines.
 * FIXES #12664 https://bugs.torproject.org/12664

 * CHANGE bridge lines created by the HTTPS distributor to not begin
   with two space characters.
 * FIXES #13123: https://bugs.torproject.org/13123

Alden S. Page authored Dec 23, 2014 and Isis Lovecruft committed Jan 08, 2015

 * CHANGE doc/HACKING.md and README.rst to mention leekspin
   installation.
 * THANKS to Alden Page for the patch.

Signed-off-by: Isis Lovecruft <isis@torproject.org>

Until we have enough bridges supporting obfs4 to actually distribute
them, and Tor Browser has obfs4 built in, I'm leaving the HTTPS dropdown
item present but disabled, and nothing will be mentioned in emails sent
by the Email Distributor until the `"obfs4"` line in
`bridgedb.strings.CURRENT_TRANSPORTS is uncommented.

In addition to the one-character bug for #12932 which was fixed in
commit 487c1b6c, there were an
additional four bugs in the legacy parser,
`bridgedb.Bridges.parseExtraInfoFile()` (which I am about to deprecate
anyway for #9380):

       # get the transport line
       if ID and line.startswith("transport "):
           fields = line[10:].split()
           # [ arglist ] field, optional
           if len(fields) >= 3:
               arglist = fields[2:]               # BUGS 1 and 2
               # parse arglist [k=v,...k=v] as argdict {k:v,...,k:v}
               argdict = {}
               for arg in arglist:
                   try: k,v = arg.split('=')      # BUG 3
                   except ValueError: continue    # BUG 4
                   argdict[k] = v
                   logging.debug("  Parsing Argument: %s: %s", k, v)

  BUG 1: This assumes the PT arguments are space-separated in the
         extrainfo descriptor. They are not; they are comma-separated.

  BUG 2: This would result in parsing the entire, comma-separated group
         of PT arguments into:

             {"key1": "a,key2=b,key3=c"}

  BUG 3: This would produce a ValueError, because there's more than one
         '=' character. (Meaning that the whole set of arguments would
         be discarded due to Bug #4.)

  BUG 4: The whole set of arguments gets discarded, without even so much
         as a log message, if there was more than one argument.

These are all bug fixes on a single commit,
4300329a, from #4568. And I'm still
deprecating the entire function anyway (for #9380) because the rest of
it is likely just as full of bugs.

 * FIXES #12932 https://bugs.torproject.org/12932

This is temporary. It is needed in order to make sure that the
regression tests for #12932 are functioning correctly. Because #12932 is
about the way PT arguments are presented to the user in the bridge
lines, to test the fixes for it, we must go through the UI (as the tests
in `lib/bridgedb/test/test_https.py` do).

 * FIXES #12932 https://bugs.torproject.org/12932