fix(arm): don't raise IndexError on malformed resourceId()/reference() calls

[ankit090701]

Description

Reported as a pipeline crash (IndexError: list index out of range, re-raised from checkov/common/parallelizer/parallel_runner.py after occurring in a worker process) that only happens when the arm framework is included in a scan.

Root cause

Two functions in checkov/arm/utils.py, both used while building the ARM dependency graph (checkov/arm/graph_builder/local_graph.py), index into a str.split(...) result without checking there are enough parts:

1. extract_resource_name_from_resource_id_func - called from _create_explicit_edge for any dependsOn entry containing the substring "resourceId(", regardless of whether the call actually has the assumed resourceId(type, name, ...) shape:

   return clean_string(resource_id.split(',')[1].split(')')[0])

   A single-argument (or otherwise malformed) resourceId() call has no comma to split on, so [1] raises IndexError.

2. extract_resource_name_from_reference_func has a mismatched guard: it checks 'resourceId' in resource_name (no trailing paren) but then unconditionally does resource_name.split('resourceId(', 1)[1] (with the paren). Any resource/variable name that merely contains "resourceId" as a substring without being immediately followed by "(" - e.g. resourceIdentifier - passes the guard but fails the split, raising IndexError instead of falling through to the plain-reference handling every other shape uses.

Fix

• Add a length check before indexing in extract_resource_name_from_resource_id_func; fall back to the raw (cleaned) string if the split doesn't have the expected shape, rather than crashing.
• Tighten extract_resource_name_from_reference_func's guard to check for 'resourceId(' (matching what the subsequent split actually looks for), so the two can never disagree.

Both changes are minimal and preserve existing behavior for every well-formed input (verified against the existing test cases).

Testing

• Added test_extract_resource_name_from_resource_id_func (a new test - the function had no dedicated tests before) covering both the documented multi-argument shape and a three-argument nested-resource shape.
• Added test_extract_resource_name_from_resource_id_func_single_arg, reproducing the crash class from this issue with a single-argument resourceId() call.
• Added test_extract_resource_name_from_reference_func_resource_id_substring_without_call, reproducing the second crash with a resource name (resourceIdentifier) that contains "resourceId" as a substring without a call.
• Verified by reverting just the checkov/arm/utils.py change: both new crash-reproducing tests fail with IndexError: list index out of range at the exact lines described above.
• Ran the full tests/arm/ suite (pytest tests/arm/): 211 passed. The only remaining issue (tests/arm/graph_builder/checks/test_yaml_policies.py) fails to even collect in my minimal Docker environment due to a missing graph_framework test-parametrization setup unrelated to this change (nothing in this PR touches graph-framework/DB-connector selection).

Fixes #7633