load data from databroker and create an in-memory H5

[leofang]

Starting from v2.9.0, h5py supports in-memory H5, which means one can allocate a segment of shared memory, within which a H5 object can live. If the frontend can do this, it will help avoid reading large data from disk in the backend.

Thoughts:

• to actually save time, we want to reuse the same H5 many times before closing the shared memory. Is this the most common use case?
• before killing the in-memory object, should we write it to disk? (I think so)

[leofang]

Caution: this will go wrong if the GUI uses multiple nodes...

[leofang]

@mrakitin this is what I mentioned the other day

[mrakitin]

For the first point, I think it makes sense to reuse.
+1 for writing to disk in case the dataset should be available later.

I'd ask @tacaswell for his thoughts as well.