feat: implement caching support for RNTuple reading

[ariostas]

This PR implements caching for RNTuple reading. I also did some useful refactoring that will help with Virtual arrays, and will also improve performance since all the buffers are pre-allocated and concatenation steps were removed.

[Copilot]

Pull Request Overview

This PR implements caching support for RNTuple reading and includes performance improvements through buffer pre-allocation and elimination of concatenation steps. The changes refactor the page reading logic to use pre-allocated buffers and introduce array caching capabilities.

Key changes:

• Added array caching support via array_cache parameter throughout the RNTuple reading pipeline
• Refactored read_col_pages and read_col_page into read_cluster_range and read_cluster_pages with pre-allocated buffers
• Modified post-processing operations to work in-place on buffers instead of creating new arrays

Reviewed Changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/uproot/models/RNTuple.py	Core implementation of caching and buffer pre-allocation refactoring
src/uproot/behaviors/RNTuple.py	Integration of array caching into the arrays() method
tests/test_1191_rntuple_fixes.py	Updated test to use renamed method and added entry_start parameter testing
tests-cuda/test_1411_rntuple_physlite_ATLAS.py	Updated comment to reflect current AOD container interpretation
tests-cuda/test_1223_more_rntuple_types.py	Simplified test filter pattern and reformatted comment

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}