Error message for nr_only flag producing empty dataset

[jjakob03]

When using the nr_only flag with not enough events in the root file or using only a part of the file, then the probability is quite high that not a single entry is returned. For example, the context & config could look like this:

st=cutax.contexts.xenonnt_fuse_sim_SR0v1_globalv12(simulation_config='sr0_dev')
st.set_config(
    {
        "path": "root/path/ambe_w_coincidence/",
        "file_name": "sim.root",
        "entry_stop": 20,
        "nr_only": True
    }
)

In this case, fuse dies dramatically with an unhelpful error message somewhere in the chunking:

numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
No implementation of function Function(<function argsort at 0x1471345ecee0>) found for signature:
 
 >>> argsort(array(int64, 3d, C))
 
There are 2 candidate implementations:
   - Of which 2 did not match due to:
   Overload of function 'argsort': File: numba/core/typing/npydecl.py: Line 368.
     With argument(s): '(array(int64, 3d, C))':
    No match.

During: resolving callee type: Function(<function argsort at 0x1471345ecee0>)
During: typing of call at .../fuse/fuse/common.py (30)


File ".../fuse/fuse/common.py", line 30:
def dynamic_chunking(data, scale, n_min):
    idx_sort = np.argsort(data)
    ^

It would be helpful if an empty array would raise an error suggesting to go for more events or explaining what happened instead.