Add Zarr benchmarks

[oruebel]

Fix #49

This PR makes the following main changes:

• Update get_object_by_name to support look-up of an object by name even if a file contains duplicate names (i.e., enforce only that the name of the object we are actually looking for is unique)
• Update core/_streaming.py to add functions for reading Zarr NWB files
• Add hdmf-zarr to the requirements
• Add network tracking benchmarks for Zarr (network_tracking_remote_file_reading.py and network_tracking_remote_slicing.py )
• Add timing benchmarks for Zarr (time_remote_file_reading.py and time_remote_slicing.py)
• Update to new BaseBenchmark class and parametrize scheme

Note: Currently we should use hdmf-zarr from the dev branch for this PR since the changes to allow us to configure to not use consolidated metadata have been merged but not released yet (or ideally even use hdfm-zarr @ git+https://github.com/hdmf-dev/hdmf-zarr.git@fix/consolidated_open_bug). However, we are planning to a release hdmf-zarr with the relevant fixes on Monday so I only listed hdmf-zarr in the dependencies, rather than setting a particular branch

[oruebel]

Note, I ran a few individual tests (which all worked), but I have not been able to rerun all the benchmarks with the additional tests yet.

[oruebel]

@CodyCBakerPhD Please take another look. I've

• updated read_zarr to not do the fallback if consolidated metadata is not present
• updated read_zarr_nwbfile to expose the mode parameter (NOTE: hdmf-zarr does the fallback check under the hood)
• updated the benchmark classes to use the new BaseBenchmark class and parametrization scheme

[CodyCBakerPhD]

Looks good to me 👍