Implementing abstract rendering and new data backends


I've been working on implementing abstract rendering using a different
data backend (i.e., not HDF5DataBackend) and with different, non
AR-framework methods (using yt's variable resolution routines) for the
resample operations. I've been successful in implementing a new
server and the abstract rendering methods for image resampling.

Before I went any further, I had a few questions that I was hoping to
get some help with:

* Presently, HDF5DataBackend is hardcoded into the server startup
phase. Is it worth a PR to make this a configurable option in the
call to configure_flask?
* The words "abstract rendering" are hardcoded as a resample
operation -- when I used other words like "yt resample" the server
never got requests sent for new data. Is it safe to just use
"abstract rendering" as the resample op with a parameter specifying
the operation as something specific to the data backend? Or could
that cause issues with other areas?
* Should I be aware of any issues with linking of data in this
process, or should the dataspace/plotspace setup working just fine to
manage all of that? (I sort of assume the latter. :slight_smile:

Thanks for any ideas!


PS Bokeh is pretty rad.