Default to VLen-UTF8 for R strings in `write_zarr_array()`

[Bisaloo]

Blocker:

• Need to support NA_character_ (see

  Rarr/R/codecs_array_bytes.R

  Line 113 in def970d

  # charToRaw() converts NA_character_ to "NA"

  )

Benefits:

• Better interoperability with python. zarr python doesn't support fixed length strings for v3
• Probably more storage efficient in most cases