Merge sen2_testing into main

[DakotaHester]

This PR contains many many changes, the most notable of which are listed below:

• New function for creating timeseries using pcxarray.prepare_timeseries()
• Switch to odc-geo for fast, Dask-compliant reproject, resampling, merge, and clip operations
• New pcxarray.get_pc_collections() function to list available Planetary Computer collections
• Enable multiprocessing functionality for prepare_data
• Automatic caching of query results
• Add new examples demonstrating new functionality on various datasets (gNATSGO, Landsat, Sentinel-2, HLS).

While this branch was originally created to test Sentinel-2 collections, it has expanded into a avenue for incorporating tons of new features into the base package.

[Copilot]

Pull Request Overview

This PR merges the sen2_testing branch into main, introducing major new features and refactors across querying, caching, and processing geospatial data.

• Refactored core processing to use odc-geo for lazy merging, reprojection, and merged timeseries support (prepare_timeseries).
• Added persistent caching for expensive operations and a new helper (get_pc_collections) to list STAC collections.
• Simplified STAC query logic with thread-based timeouts, retry/backoff, and updated I/O functions.

Reviewed Changes

Copilot reviewed 8 out of 13 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
src/pcxarray/utils.py	Added caching decorator, renamed progress param, updated utilities
src/pcxarray/query.py	Replaced multiprocessing with threading, added retries & caching, introduced get_pc_collections
src/pcxarray/processing.py	New processing module: lazy_merge_arrays, refactored prepare_data, added prepare_timeseries
src/pcxarray/io.py	Enhanced raster loading (load_from_url), single-item reading logic
src/pcxarray/cache.py	New caching module using joblib.Memory
src/pcxarray/init.py	Updated public API exports
pyproject.toml	Added dependencies: joblib, odc-geo, bottleneck
README.md	Expanded overview, usage examples, and documentation updates

Comments suppressed due to low confidence (6)

src/pcxarray/io.py:59

• The docstring references a timeout parameter, but the function signature does not include timeout. Either remove or add the parameter.

    timeout : float, default 60.0

src/pcxarray/processing.py:496

• [nitpick] New prepare_timeseries API introduces complex behavior but lacks unit tests to validate grouping, parallel execution, and output structure.

def prepare_timeseries(

src/pcxarray/utils.py:120

• The function uses requests.get but import requests is missing at the top of this file, causing a NameError.

        with requests.get(url, stream=True, verify=verify) as r:

src/pcxarray/processing.py:250

• Iterating for item in items_full_overlap loops over column names, not rows. To try each feature record, use .iterrows() or .itertuples() and pass the correct row to read_single_item.

        for item in items_full_overlap:

src/pcxarray/processing.py:89

• This check will raise an error when all three arguments are None, but the intention is to auto‐compute them when all are missing. Adjust the condition to allow sum(...) == 3 to proceed with automatic determination.

        if sum([geom is None, crs is None, resolution is None]) > 1:

src/pcxarray/processing.py:116

• [nitpick] Typo in comment: "rekli=ommon" should be "common".

    # reproject all arrays to the rekli=ommon geobox