Should we write to disk HDF5-settings-related LGDO attributes?

[gipert]

We drop the compression and hdf5_settings attributes before writing LGDO to disk, at the moment. This is because their names are specific to this Python implementation (h5py.create_dataset()) at the moment, so we should not write them to disk as they are. They are also redundant, since these settings can be also deduced from the HDF5 metadata (waveform compression is instead documented with attributes specifically defined in our data format specification).

The downside is that if a user wants to load a dataset in memory and dump it again to disk after some minor modifications, they will need to manually specify HDF5/compression settings again, which is quite annoying.

[jasondet]

we could detect the setting on lh5.store.read() and set the corresponding attributes at that time?