- treat empty input buffers as heart beats, add code to forward 0-length
  buffers out each source pad

- edit property help text
- don't strip whitespace from channel names.  they can't have whitespace,
so this shouldn't matter, but removing it might mask bugs in other code so
it's probably better to not do it.

allows the user to select a limited list of channels to demultiplex.
hopefully this allows the element to work practically with level 0 frame
files

- edit error messages in the pad-linked handler to be associated with the pad instead of the element

- oops.  thought the sink pad was forwarding events but it wasn't.  eos
  wasn't getting through, in particular, which prevented pipelines from
  stopping.

- simplify and increase the speed of the pad state object look-up by using the "element_private" pointer of the GstPad objects.  this patch creates a memory leak, but it's not triggered in normal use and will go away eventually

- src pads that got peers attached to them after eos was been received used to get a belated newsegment event pushed out.  don't do that

- disable framecpp_igwdparse element:  required base class not available in 0.10.32 on clusters.  will re-enable when I can come up with a configure test to check for required version.

this element implements a gstreamer parser for IGWD frame files.  i'm
pretty sure it's totally borken:  it only works for files in which 4-byte
ints are 4 bytes, etc., and I have no idea how to identify an end-of-file
marker robustly so I just copied the "class" code I found in the hex dump
of a frame file on my own computer.  anyway, it's enough to let me run
totem on a frame file and have it actually do something! :-).
unfortunately it just sits there because it thinks it needs to wait several
hundred million seconds before rendering something, but that'll get sorted
out eventually.

this work includes the addition of a "parsed=true" field to the caps
claimed by the framecpp_channeldemux and gds_lvshmsrc elements.  meanwhile
the igwd typefinder reports "parsed=false", which induces gstreamer's
autoplugger to insert the new parser element between generic source
elements and the channel demuxer.

- don't use the global FrameCPP default namespace;  makes it easier for me to understand where various symbols are coming from.

- fix a namespace problem in the full-file CRC code.  g++ 4.4 does not
  complain about it, but code is uncompilable with g++ 4.6 without this
  patch.

- adjust typefind function rank and name
- adjust framecpp_channeldemux rank

- add typefind support.  teaches gstreamer (e.g. filesrc element) to
  autodetect IGWD frame files.  tested with some S5 LHO frame files.

This reverts commit 42adbf35.

- allow file checksums to be enabled/disabled
- allow bad files to be non-fatal

- fix a couple of typos in channeldemux

- log new caps for pads