…out and tensor creation

- Add ShardMode enum to specify shard shape in shard spec as either physical or logical
  * ShardMode::PHYSICAL: This is current behaviour that we will deprecate!
    ** It is less expressive than using shard shape as logical (ie. must be tile aligned for TILE layout etc...)
    ** It fundamentally operates on padded shape and is confusing and incompatible with logical shape
  * ShardMode::LOGICAL: Shard shape cuts 2D logical shape and each shard is aligned after
    ** Without alignment restrictions, you can cut 2D logical shape more arbitrarily
    ** Existing sharding can be switched over to this entirely (just need codeowners to help out and flip...)
  * Default everywhere will be ShardMode::PHYSICAL with a warning message
- Switch tests/ttnn/unit_tests/operations/test_paged_update_cache.py to use logical shard shape as an example
  * Introduce tensor.logical_volume() (as opposed to tensor.volume() which returns physical volume based on padded shape)
  * TODO: Rename volume() -> physical_volume() and logical_volume() -> volume()
- Add new c++ tests to test tensor creation with logical shard shape + alignment
  * IMPORTANT: Need to update host data manipulation to be aware of new logical sharding for use from python!

To support these changes, some changes to TensorLayout:
- Make private TensorLayout constructor with alignment public with these changes:
  * legacyShapeToAlignment will try to return 2D alignment if possible (ie. only padding on height/width)
    ** Goal is to transition alignment to be 2D only if we remove poor use cases of padding on non-height/width dims
  * legacyShapeToAlignment is only expected to be used for ShardMode::PHYSICAL and uses default alignment for sharded tensors
    ** Before interleaved or sharded will just use padded shape for alignment
    ** One exception is for row major sharded tensors where we use shard width if shape is padded;
       otherwise, we only take shard width for BLOCK/WIDTH sharded cases and original physical shape for HEIGHT sharded
  * legacyShapeToAlignment (and alignment in general) will work iff there is only padding on height and/or width
    ** IMPORTANT: This means we are expecting tensors with arbitrary padding along non-height/width to be interleaved only!
- If ShardMode::LOGICAL:
  * In TensorLayout::compute_shard_spec_buffer, calculate physical shard shape based on shard shape + alignment
  * In TensorLayout::compute_physical_shape, calculate physical shape based on number of logical shards
- Clean up handling of sharded tensors and error messages in ttnn/cpp/ttnn/tensor/layout/page_config.cpp
- Add Size constructor for std::array<uint32_t, 2>