[PATCH v2 0/6] clean up author parsing

This is a re-roll of my series from June:

  http://thread.gmane.org/gmane.comp.version-control.git/252055

The goal is to clean up some ad-hoc parsing code. Besides reusing code, this fixes some memory leaks, fixes some bad pointer math in determine_author_info, and protects us against getenv() implementations that use a static buffer.

This iteration incorporates feedback from the Eri[ck]s (particularly the last patch, which is much simplified). In addition, I dropped the patch that converted split_ident's pairs of pointers into a struct. It caused quite a bit of noise through the code base, and only saved us a few lines in the end (actually no lines -- it just made a few lines shorter).

The original was also based on another pending series in 'next', which has since graduated to master. This one is based directly on master.

  [1/6]: commit: provide a function to find a header in a buffer
  [2/6]: record_author_info: fix memory leak on malformed commit
  [3/6]: record_author_info: use find_commit_header
  [4/6]: use strbufs in date functions
  [5/6]: determine_author_info: reuse parsing functions
  [6/6]: determine_author_info: copy getenv output

-Peff