Validate indexes for CONTROL data

[takluyver]

Previously, we were only checking the indexes for INSTRUMENT data, not CONTROL. I think we may have assumed that CONTROL was unlikely to go wrong. On #187, we found a run where there is a problem - CONTROL data missing, INDEX still records it as though it was there - which extra-data-validate did not pick up.

This extends the same checks we do on INSTRUMENT data to CONTROL, with one difference. We only record the "Index referring to data outside dataset" problem once per source & per file, rather than for every key. When I recorded it for every key, I got a list of some 27000 problems with the run. Once per source should be plenty to point out the problem, and then you can use h5glance to find more detail if necessary.

Checking run directory: /gpfs/exfel/exp/SPB/201901/p002316/raw/r0174
Progress: 207/207 [##################################################]
84 problems in 7 files (last: RAW-R0174-DA02-S00002.h5)
Validation failed! 84 problems:

Index referring to data (11) outside dataset (0)
  dataset: CONTROL/SPB_IRU_AGIPD1M1/CTRL/MC1/module2/dBoardTemperature2/timestamp
  file: /gpfs/exfel/exp/SPB/201901/p002316/raw/r0174/RAW-R0174-DA02-S00006.h5

Index referring to data (11) outside dataset (0)
  dataset: CONTROL/SPB_IRU_AGIPD1M/TSENS/Q2_T_BLOCK/force/timestamp
  file: /gpfs/exfel/exp/SPB/201901/p002316/raw/r0174/RAW-R0174-DA02-S00006.h5

...

[tmichela]

I wonder if we should try to summarize similar problems to avoid returning N times the same output with just a different source name, ex:

Index referring to data (11) outside dataset (0) for 82 sources, example:
  dataset: CONTROL/SPB_IRU_AGIPD1M1/CTRL/MC1/module2/dBoardTemperature2/timestamp
  file: /gpfs/exfel/exp/SPB/201901/p002316/raw/r0174/RAW-R0174-DA02-S00006.h5

and add a verbose argument.

But that's a different issue.

LGTM

[takluyver]

Thanks - I've opened #190 for summarising the issues.