More options for eval engine

[jpn--]

This pull request introduces several enhancements and new features to the sharrow library, focusing on improving data handling and expression evaluation capabilities. Key changes include the addition of new methods for evaluating expressions, improvements to data handling, and the introduction of a new class to manage subspaces within a data tree.

Enhancements to Expression Evaluation:

• Added new parameters to the get_expr method to support more flexible evaluation, including dtype and with_coords. (sharrow/relationships.py)
• Introduced eval and eval_many methods to evaluate single or multiple expressions with various engines and return results as DataArray or Dataset. (sharrow/relationships.py)

Improvements to Data Handling:

• Modified the __getitem__ method to check for cached evaluations before attempting to retrieve items from the data tree. (sharrow/relationships.py)
• Enhanced broadcasting logic in the get method to ensure results are transposed according to the root dimensions or specified dimension order. (sharrow/relationships.py)
• Added handling for top dimension names in _getitem to ensure consistent dimension labeling across datasets. (sharrow/relationships.py)

New Class for Subspace Management:

• Introduced the DataTreeBranch class to facilitate access to variables in subspaces of a data tree, supporting the use of python and numexpr engines for dotted data. (sharrow/tree_branch.py)

Additional Changes:

• Updated imports to include Mapping, Sequence, and Literal from the collections.abc and typing modules. (sharrow/relationships.py)
• Added new tests to verify the functionality of the eval and eval_many methods with different engines and digitization settings. (sharrow/tests/test_relationships.py)