BEP042 has been accepted/merged, which introduces new possibilties for coordsystem.json files. Specifically:

• there may be multiple coordsystem.json files, each defining coordinates for a different part of the body
• there may be "nested" coordsystem.json files, e.g., one specifying a anatomical-landmark-relative system describing device location, and another defining a device-internal coordsystem describing the relative positions of individual sensor contacts.
• the coordinate systems themselves may have user-chosen arbitrary names like "Grid1", "LowerLeg", etc (e.g., there is no pre-defined set of coordinate systems like "CapTrak" etc)

What has not changed is that (at least one) coordsystem.json is still required if electrodes.tsv exists.

We need to decide how to handle these changes, in a few different ways:

1. what to do on read. read_raw_bids will look for coordsystem.json if electrodes.tsv is defined; currently it fails if there are 2 coordsystem.json files
2. what to do on write. Should we, e.g., provide a helper func to let users create DigMontage objects that they can pass in to write_raw_bids? Should we just encourage them to hand-write the coordsystem.json files? Should we provide a helper func to aid in writing the files, but not bother with the DigMontage object?

Note that it is possible for EMG datasets to have Polhemus-style digitized electrode locations, but this is far from commonplace. So far I haven't seen a dataset that has that (@neuromechanist are you aware of one?) but ideally we should support that case too.

Originally posted by @drammock in #1129 (comment)