iOS / asi-http-request

1) Fix a problem where a temporary request body would be discarded when retrying on a new connection
2) By default, all requests using POST, PUT or having a body at the time requestMethod is set will NOT use a persistent connection.
You can override this by setting requestMethod manually, then setting shouldAttemptPersistentConnection to YES

closes gh-94

Updated SSL test to handle updated error message

All requests now store a single expiry timestamp constructed from either secondsToCache, a cache-control: max-age header, or an expires header (in order of precedence)
This is used for all expiry checks, and should speed up checking if cached data is stale because it elimiates the need for NSDateFormatter or NSScanner when reading from the cache
(closes gh-173)

Internal tweaks to remove a bit of duplication

closes gh-164

Add trivial test for base64
closes gh-165

Many thanks to Garrett Bjerkhoel for the code and patience in getting this merged (I've slightly adapted his original take on this)

Conflicts:

	Classes/ASIHTTPRequest.m

credentials

…request, as long as you explictly set authenticationScheme to kCFHTTPAuthenticationSchemeBasic

Add readme for GHUnit

* GHUnit is no longer included as a git submodule. Instead, build scripts are used to fetch a pre-compiled GHUnit framework if one does not exist in External/GHUnit.
* iPhone sample has moved to using the framework rather than the static library for GHUnit
* Fiddle with conditional compilation for NSXMLParserDelegate
* Tweak the build phases to move running the scripts earlier
* Fix a clang warning that crept in as part of the recent changes to behaviour for fetching PAC files
* Remove legacy clang.sh script
* Move build scripts into their own folder

Conflicts:
	Classes/ASIHTTPRequest.m

…TLM (non-proxy) auth stuff
Tweaks to clean things up a bit

Previously the loop terminated when the output buffer was not completely filled by the previous call. Most of the time this is correct, because if the output buffer was completely filled, there is probably more to come. However, in the situation that the output fits perfectly in the output buffer, the buffer will be completely filled, yet there is nothing left to process.  In this case we would run through the loop again calling inflate() / deflate() which would return Z_BUF_ERROR. This would manifest itself as the request (needlessly) failing.

Rather than this we should simply terminate the loop when there is no more input left to process.

# Please enter the commit message for your changes. Lines starting
# with '#' will be ignored, and an empty message aborts the commit.
#
# Committer: Alex Deem <alexdeem@Alex-MacBook-Pro.local>
#
# On branch master
# Changes to be committed:
#   (use "git reset HEAD <file>..." to unstage)
#
#	modified:   Classes/ASIDataCompressor.m
#	modified:   Classes/ASIDataDecompressor.m
#

…ese would ever have leaked, but clang might warn about this.

…ting it back to UTF-8
Add tests for ASIWebPageRequest encoding
Also, strip any gzip content-encoding header, since the parsed response will not be deflated (cache does this anyway, but we can't guarantee we'll have one)