fix: replace deprecated JSONPropertiesContainer with AbstractPropContainer

[nanookclaw]

Replaces deprecated JSONPropertiesContainer with AbstractPropContainer in two files, per the deprecation warnings on AbstractPropertiesContainer, YAMLPropertiesContainer, and JSONPropertiesContainer in prop_manager.py.

Changes

dataproc.py (7 call sites):

• Import: AbstractPropContainer from spark_rapids_tools.utils.propmanager; kept is_valid_gpu_device from prop_manager
• JSONPropertiesContainer(prop_arg=X, file_load=False) → AbstractPropContainer(props=X) at 5 locations (pricing config, instance descriptions, worker props, master props)
• Type annotation updates at pricing_config parameter and _set_zone_from_props signature

cluster_inference.py (1 call site):

• Import swap to AbstractPropContainer
• JSONPropertiesContainer(cluster_conf, file_load=False) → AbstractPropContainer(props=cluster_conf)

All changes follow the AbstractPropContainer API: props= for dict wrapping, matching get_value/get_value_silent method signatures preserved from the parent class.

Closes #1898

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the previously reported JSON-boundary regressions are fixed and no new actionable issue remains.

Summary

• Dataproc instance descriptions accept either serialized JSON or decoded lists.
• Cluster inference accepts either rendered JSON text or decoded dictionaries.
• Regression tests cover both representations at each corrected boundary.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[CLI or rendered template output] --> B{Value is a string?}
    B -->|Yes| C[json.loads]
    B -->|No| D[Use decoded value]
    C --> E[AbstractPropContainer]
    D --> E
    E --> F[Dataproc processing or cluster loading]

Loading

_{Reviews (2) · Last reviewed commit: "fix: decode serialized JSON configuratio..."}

[nanookclaw]

Updated at commit cafb5be586a8dbae862f2ced016f4a54fa2c415e.

The change decodes serialized JSON at the Dataproc instance-description and cluster-inference configuration boundaries while preserving already-decoded inputs. Added focused regression coverage for both forms at each boundary. The Dataproc test exercises DataprocCMDDriver._process_instance_description.

Local verification under Python 3.12.3: focused test_json_boundaries.py — 4 passed; py_compile — passed; git diff --check — passed. The commit includes the required DCO sign-off.