DANDI validation rejects Zarr v3 string arrays

I converted an IBL session while testing [hdmf-zarr's v3 migration](https://github.com/hdmf-dev/hdmf-zarr/pull/325) and tried uploading it to [sandbox dandiset 218696](https://sandbox.dandiarchive.org/dandiset/218696). `dandi upload` stopped at validation:

```text
[zarr.tensorstore_cannot_open] .../identifier
Error opening Zarr array with tensorstore
```

All 113 string arrays fail this check. The other 176 numeric and boolean arrays pass. Opening `identifier` directly with TensorStore gives:

```text
Error parsing object member "data_type": string data type is not one of the supported data types
```

This is enough to reproduce the validation error:

```python
import numpy as np
import zarr

group = zarr.open_group("strings.zarr", mode="w", zarr_format=3)
array = group.create_array("identifier", shape=(1,), dtype=np.dtypes.StringDType())
array[0] = "hello"
```

Then run `dandi validate strings.zarr` (outside a dandiset it also reports `DANDI.NO_DANDISET_FOUND`).

I used DANDI 0.78.0, TensorStore 0.1.85 and Zarr 3.3.0, overriding DANDI's `zarr<=3.1.5` pin. I also checked an array written with Zarr 3.1.5 and TensorStore rejects it with the same dtype error.

I think `_ts_validate_zarr3` is rejecting an encoding that Zarr-Python can read. Its docstring already says it should go away after the Zarr-Python 3 migration. Could we use Zarr-Python to validate these arrays? What do you think?

Start at the `_ts_validate_zarr3` entry point and reproduce the failure with the provided Zarr v3 string-array example, then run `dandi validate strings.zarr`. Trace how the validation handles the reported string dtype and compare it with the arrays Zarr-Python can read. Done means string arrays validate successfully while the existing numeric and boolean array checks continue to pass.

Written by the indexing model from the issue text.