Change imagemap.sh to use updatethumbnails instead.

Change photofokus.el to use updatethumbnails instead.

Added the new -i option to updatethumbnails, which photofokus.el needs
to be able to rotate an image without making a new commit.

The embedded thumbnails are read using jhead -st.  That code path is
however only used if only small thumbnails are created.

The new options "-c camera", "-s size" and "-t synctime" can be used
in any combination to limit the files that are processed.

The new script is a lot more clever, so when nothing has changed is
exits in less than a second.  It is also roughly 10% faster on scaling
the images, since it only calls djpeg once per image.

All the hotsync scripts now create a "complete" file once all images
are loaded.  updatethumbnails.py refuses to process a directory until
that file exists, so that it doesn't create its own "directory
completely processed" file before all files are loaded from the
camera.

The "directory completely processed" file is named "sha1" and contains
the Git commit ID of /photos/metadata.  If it changes, the script may
have to recompute the tumbnails due to changed image rotations.
Otherwise it can really quickly skip the directory.

If a pre-scaled fullscreen-sized image is available, use that, so that
slidegtk does not have to resize the image.  This makes the viewer
program more responsive.

Instead of doing (elt pf-images n) up to three times per image, do it
just once and save the result in a local variable.  If pf-images is
long, this can speed up things quite a bit.

This is more robust than remembering the current buffer and doing
(set-buffer old) when done.  Affected functions:

 pf-read-file
 pf-write-file
 pf-read-property-file
 pf-write-property-file
 pf-load-one-syncdir

The cache is created (if needed) when all images are loaded.  It is
removed if the user edits any metadata of an image.

It used several bad coding practices, that made it very slow when
there were many images.  Now it is a lot faster.

Also, don't require the pf-tag-index to store the images in sorted
order.  The code doesn't seem to require that, anyhow.

The progress implementation broke image creation.

Try each in turn until a non-empty one is found.

It combines pf-find-all-images with pf-find-all-tags.

New scripts: d5000-hotsync.sh, d5000-remove.sh.

Added the "magic" class to photofokus.el and updatethumbnails.sh.

New scripts: magic-hotsync.sh.

Added the "magic" class to photofokus.el and updatethumbnails.sh.

imagemap.sh: Only display the removal button if a removalkey file was found.

New files: s520-hotsync.sh, s520-remove.sh.

Apparently, the files are no longer word readable, which we want them to be.

Adjusted c2100uz-hotsync.sh, imagemap.sh, mkscalecache.sh and photofokus.el.

(pf-slideshow): Add one to the last image in the interactive
  specification, so that the entire range is included in the
  slideshow.

(pf-update-n): Use dolist instead of repeatedly using length and elt
  on the list of images.