[Bug] Misleading DoesNotExistErrors in open calls

[ryan-williams]

1. Experiment.open raises DoesNotExistError when passed a URI to a different type (e.g. a SparseNDArray URI within an Experiment)

2. All .opens (including tiledbsoma.open) raise DoesNotExistError if AWS_{DEFAULT_,}REGION is not set (though e.g. aws s3 ls figures it out).

   Occurs even if ~/.aws/config contains:

   [default]
   region = us-west-2
   

A clearer error message would help, in both cases.

Original description follows:

---

import tiledbsoma
uri = 's3://cellxgene-census-public-us-west-2/cell-census/2024-07-01/soma/census_data/homo_sapiens/ms/RNA/X/raw'
tiledbsoma.open(uri)             # ⚠️ Passes on GHA, fails locally on M3 Macbook and Ubuntu EC2 instance
tiledbsoma.Experiment.open(uri)  # ❌ Fails locally and on GHA

# ✅ OK; URI is resolvable with AWS environment configs
aws s3 ls s3://cellxgene-census-public-us-west-2/cell-census/2024-07-01/soma/census_data/homo_sapiens/ms/RNA/X/raw

Here's an example GHA showing the Experiment.open failure on {1.14.5,1.15.0rc4} x py{3.9.19,3.11.8} x ubuntu-latest.

It seems that clib.SOMAObject.open can return None for a variety of reasons, here, but we always just raise DoesNotExistError.

Versions (please complete the following information):

• TileDB-SOMA version: 1.14.5, 1.15.0rc4
• Language and language version: Python 3.9.19, 3.11.8
• OS: MacOS, Ubuntu

From one of the GHA failures:

tiledbsoma.__version__              1.[15](https://github.com/ryan-williams/tiledb-scratch/actions/runs/12015718039/job/33494374982#step:5:16).0rc4
TileDB core version (libtiledbsoma) 2.27.0
python version                      3.11.8.final.0
OS version                          Linux 6.5.0-1025-azure

[ryan-williams]

@johnkerl ID'd 2 issues I was hitting:

1. Experiment.open was failing because I was trying to open a SparseNDArray
2. Both were failing locally because I didn't have $AWS_REGION set

Would be great to improve on the DoesNotExistError for both cases.

[johnkerl]

Note we need to implement this wired all the way through C++/Python/R -- I'm adding myself as the R assignee

[johnkerl]

Notes about the R side:
#3402 (review)

[johnkerl]

I believe #3402 resolves this -- if I'm mistaken, please let me know and I'll happily re-open.