Make upload.py send svn cp'd and svn mv'd files correctly.

This resolves http://code.google.com/p/rietveld/issues/detail?id=82.

Note that the patch will look correct in Rietveld, but won't have the same effect when downloaded via the "Download raw patch set" link; I could conceivably have upload.py send up the entire file, but that still wouldn't produce the same effect in the patched working copy (specifically, it wouldn't fix up SVN's internal metadata).

[Collin Winter, 2009-04-10 21:42:14]

Hey Guido, please take a look at this patch to upload.py. Let me know how you
want to handle the "Download raw patch set" issue. I'm not sure what the right
answer is there.

[GvR, 2009-04-10 22:53:56]

Sorry Collin, I'm all out of round tuits this week. Maybe Andi can have a look?

[Andi Albrecht, 2009-04-11 12:42:40]

IMO when downloading the diff from codereview it should apply as clean as
possible to give reviewers a chance to patch it in and test it locally.

As the diff always is in unified diff format VCS specific history will be lost,
at least for SVN. I'd prefer to have the base file uploaded for copied sources.
I think, it's up to the contributor to communicate that a certain file should be
copied when a patch goes in.

http://codereview.appspot.com/40050/diff/1/2
File static/upload.py (right):

http://codereview.appspot.com/40050/diff/1/2#newcode806
Line 806: data = RunShell(diff_cmd)
I think you should use RunShellWithReturnCode here or set the "silent_ok" flag
to True. If the changeset only contains a copied file from a remote location,
the script will stop here with "No output from ['svn', 'diff']". It only works,
if the changeset has other changes too.

http://codereview.appspot.com/40050/diff/1/2#newcode830
Line 830: data = RunShell(status_cmd)
Again, this raises an ErrorExit if "svn status" returns no output. IMO it's
better to raise the "No valid patches found..." exception in line 861.