Collection of small improvements from the HTTP Archive work

[pmeenan]

• Makes the GZip level configurable: Allows for space/cpu tradeoffs for the test data stored by the php (not the gzip fiules from the agent). Dropping the gzip level to 1 cut the time in half to write the test job file on the HA server. Defaults to 6 which is the original hard-coded value.
• Added a flag (lighthouse=0) to the HAR export API to allow for disabling the including of lighthouse data in the HAR when it isn't needed.
• Moved the daily test counter incrementing into APC using atomic incrementing instead of a locked filesystem file. This reduces the contention when lots of tests are being submitted concurrently and reduces the overhead for each test.
• Fixed an exception that would be thrown in the HAR generation code if the browser data wasn't populated.
• Added a check to not re-generate the page_data on the server if it was already processed on the agent.

[jefflembeck]

This checks out to me and looks great.