issues with data chunks

[frostschutz]

The readme says

• Works with chunks of a data - no need to buffer the whole request

but it doesn't seem that simple. If a header happens to be in a chunk boundary, you get double callback calls with half a header field name in each. So instead of blindly passing whatever data was available I have to do my own buffering, or track state and buffer whatever I get in the callbacks.

The part_data callback also is very noisy, it gets tons of calls for 1 byte or even 0 byte even if the data chunk the parser currently works on is considerably larger and a larger data block could be passed along consecutively w/o additional buffering.