multipart/form-data parsing restored

[Massimo Manghi <ma...@unipr.it>]

Hi 

I gave up trying to figure out the purpose of
the changes made on apache_multipart_buffer.c
and in order to have multipart/form-data messages
parsed correctly I (almost) reverted it back to 
the code for apache-1. 

IMO the level of maturity of the code in 
apache_multipart_buffer.c is not satisfactory
and (as Valery wanted to do) should evolve
(read: rewritten) to exploit libapreq2 parsers. 
I can take up the task but I have to put it
off for a while now

Here is the changelog entry for today's patches 
(patches are attached to the message)

  * INSTALL: instructions for builing Rivet expanded with
  examples taken from real builds for Debian and Slackware.

  * src/apache-2/apache_request.[c|h]: removed TODO comments
  concerning ap_hard_timeout and ap_kill_timeout as these functions
  doesn't not exist anymore in the apache 2.x api.
  Added the Content-type as a new argumento to
  ApacheRequest_parse_multipart for parts boundary determination.
  The correctness of the content type has to be done before calling
  in order to do the check only once. Commented declarations of variables
  not used. The 'tempname' string is allocated now from the apache pool.
  Changed template for temporary file name generation.
  If apr_file_mktemp fails a more descriptive error is logged.

  * src/apache-2/apache_multipart_buffer.[c|h]: removed line in
  multipart_buffer_read that lead the buffer pointer to fail
  following the parsing process. Input buffer size changed to a
  larger and more common size of 8192 bytes

-- Massimo