Preserve slice projections after array indexes

[Eric3-jp]

For data = {'foo': [[{'a': 1}, {'a': 2}]]}, the expression foo[0][:2].a currently returns None instead of [1, 2]. A slice following an array index bypasses projection construction, so subsequent expressions are evaluated against the entire sliced list instead of its elements. foo[0] | [:2].a already produces the expected result, consistent with the documented slice projection behavior.

Exclude slices from the optimization that combines consecutive indexes, allowing them to use the existing _project_if_slice path. Consecutive scalar indexes retain the optimization.

Add 13 compliance cases covering field/root expressions, positive and negative indexes, repeated indexes, reverse and empty slices, missing fields, subsequent array indexes, pipe boundaries, non-array inputs, and null omission. The changelog is updated.

Validation on CPython 3.11.16 under Ubuntu 24.04, using the repository's test dependency versions:

• Before the parser change, the slice compliance file had 8 failures and 46 passes; all failures were among the new cases.
• Afterward, python -m pytest tests -q --cov=jmespath --cov-report=term-missing: 1004 passed, 1 skipped, 99% overall coverage.
• git diff --check: passed.

The full Python/platform CI matrix was not run locally.

Developed with assistance from OpenAI Codex.