Cut data point upload batch size to 25.

Cut data point upload batch size to 25.
Some bots have lower speed sending http data out, and this will make it less likely to timeout.
Did not try 50 because once we switch to using Chrome dashboard app, extra work needs to be done for each data point which will take longer. Since appengine has the fixed 30s timeout policy, there's not too much we can do unless we convince Chrome to switch to other solutions.
Current Skia data on our app instance: https://skiadash.appspot.com/report
Committed: https://code.google.com/p/skia/source/detail?r=8133

[benchen, 2013-03-13 17:12:23]

First try on batch size adjustment. Already temporarily changed the app side to
prevent some other potential problems.

[borenet, 2013-03-13 18:26:23]

On 2013/03/13 17:12:23, benchen wrote:
> First try on batch size adjustment. Already temporarily changed the app side
to
> prevent some other potential problems.

LGTM.  This is still disabled on the bots.  Once things start to cycle green,
I'll enable on one platform at a time to make sure this helps.