Empty hits list error still happens after 0.9.3

[svenrdz]

From @orlando-code in #119

When no hits are returned by context's search(), perhaps due to a time-out or incorrect search string, context.py's _distribute_hits_impl attempts to index an empty list. This fails with the following error:

...
  File "../python3.14/site-packages/esgpull/context.py", line 268, in _distribute_hits
    offsets = _distribute_hits_impl(hits, offset)
  File "../python3.14/site-packages/esgpull/context.py", line 250, in _distribute_hits_impl
    accs[i] += steps[i]
    ~~~~^^^
IndexError: list index out of range
IndexError: list index out of range

I ran into the above error when running searches with the bridge index node (index_node = "esgf-node.ornl.gov/esgf-1-5-bridge").

It's not a timeout issue (I made the http_timeout values very large; and the error takes the same length of time to appear). Logger gives this readout:

[2025-12-02 13:46:16]  INFO      esgpull
facets=('project:"CMIP6"', 'activity_id:"ScenarioMIP"', 'experiment_id:"ssp126"', 'institution_id:"CNRM-CERFACS"', 'source_id:"CNRM-ESM2-1"', 'variable_id:"ph"')

...

[2025-12-02 13:46:16]  DEBUG     httpcore.connection
connect_tcp.started host='esgf-node.ornl.gov' port=443 local_address=None timeout=100000 socket_options=None

[2025-12-02 13:46:16]  DEBUG     httpcore.connection
connect_tcp.complete return_value=<httpcore._backends.anyio.AnyIOStream object at 0x79d652379940>

[2025-12-02 13:46:16]  DEBUG     httpcore.connection
start_tls.started ssl_context=<ssl.SSLContext object at 0x79d654b679b0> server_hostname='esgf-node.ornl.gov' timeout=100000

[2025-12-02 13:46:51]  DEBUG     httpcore.connection
start_tls.failed exception=ConnectError(EndOfStream())

[2025-12-02 13:46:51]  DEBUG     esgpull
fetch (1 sub-exception)
NoneType: None
[2025-12-02 13:46:51]  DEBUG     esgpull

NoneType: None
[2025-12-02 13:46:51]  DEBUG     root
Locals:
{'hits': [], 'max_hits': 0, 'i': 0, 'total': 0, 'N': 0, 'accs': [], 'result': [], 'steps': []}

---

There was a fix for this error in #110 but it obviously did not fix the issue.