they are quite private and should (usually) be singeltons.
- updated all instances to use UPPPER_CASE format, like CONTENT_TYPE
- propigated these changes to tests and to fileapp.py

- renamed Expires.time to Expires.parse for consistency
- updated FileApp/DataApp to return 206 on Partial Content
- all HttpHeader(environ) return strings (empty string when not found)
  so that checks like 'if header-part in HttpHeader(collection)'
  works without having to check for None
- updated FileApp to use Range header (instead of having its own copy)

  the corresponding header is not found (hopefully this redundance can
  be removed from WSGI)

- fixed bug in content-disposition
- fixed bug in cache-control

next-up: Content-Range, Range headers

- added DateHeader base class for Expires, If-Modified-Since, etc.
- added apply method for updates /w side-effects

- added delete and update methods
- added regression tests

  of outgoing headers and capitalizes them to exactly match the RFC
  for extremely dumb user agents (which exist... unfortunately)
  that don't do case-insensitive matches and assume Camel-Case
- added ResponseHeaderWrapper to use python properties to
  access headers which don't use multiple-entities (like Set-Cookie)
- updated comment to header_value to explain why , separator
  works in most cases (but not all)

  accidently used as a key in an WSGI  or header tuple
- distinguishing between multi-value (but single-line) and
  multi-entry headers such as WWW-Authenticate and Set-Cookie
  as discussed on the Web-Sig
- removed 'dict' like stuff from environ wrapper so that it
  isn't accidently used as an environ replacement