Describe the bug
The string matching routine suggested in the documentation via "data_%T.h5" fails for mixed output wich filter strings as commonly used in PIConGPU.

To Reproduce
Assume you have data of the form simData_100.h5 and additionally filtered data in the form simData_filter_100.h5 in the same directory. While the suggested placeholder %T should select iterations only, it is confused with the additional string beta_ and throws the following run time error:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-36-4ed3e1a89505> in <module>
      2 series = api.Series(
      3     path + "simData_%T.h5",
----> 4     api.Access_Type.read_only)

RuntimeError: Unsupported dataset class

Python:

import openpmd_api
path = "/scratch/ws/<...>/runs_LWFA_PWFA/069_<...>/simOutput/h5/"
series = api.Series(
    path + "simData_%T.h5", 
    api.Access_Type.read_only)

Expected behavior
Intuitively a expected %T to only match numbers (with and without leading zeros) and thus ignore additional letters (as used by filter names in PIConGPU).

Software Environment

• version of openPMD-api: 0.10.0-alpha
• installed openPMD-api via: conda-forge
• operating system: Red Hat Enterprise Linux Server release 7.4 (Maipo)
• machine: taurus ZIH TU Dresden
• name and version of Python implementation: Python 3.6.9 | Anaconda, Inc.
• version of HDF5: 1.10.5
• version of ADIOS1: 1.13.1
• name and version of MPI: no MPI used

Additional context
This effects use of all filtered openPMD data in PIConGPU.