Migration to the Finch Julia backend

To switch to the Finch Julia backend, set the environment variable SPARSE_BACKEND="Finch", then continue using.

While this is largely compatible with the Array API, support for some functions may not be present, and API compatibility isn't strictly preserved with the default (Numba) backend.

However, the new backend has a large performance benefit over the default backend. Below, you will find a table of common invocations, with their equivalents in the Finch Julia backend. The most common change is a standard API for construction of arrays.

Numba Backend (`SPARSE_BACKEND="Numba"`)	Finch Julia Backend (`SPARSE_BACKEND="Finch"`)	Notes
`sparse.COO.from_numpy(arr, fill_value=fv)` `sparse.COO.from_scipy(arr)` `sparse.COO(x)`	`sparse.asarray(x, format="coo", [fill_value=fv])`	Doesn't support pulling out individual arrays
`sparse.GCXS.from_numpy(arr, fill_value=fv)` `sparse.GCXS.from_scipy(arr)` `sparse.GCXS(x)`	`sparse.asarray(x, format="csf", [fill_value=fv])`	Format might not be a 1:1 match
`sparse.DOK.from_numpy(arr, fill_value=fv)` `sparse.DOK.from_scipy(arr)` `sparse.DOK(x)`	`sparse.asarray(x, format="dok", [fill_value=fv])`	Format might not be a 1:1 match

Most things work as expected, with the following exceptions, which aren't defined yet for Finch:

sparse.broadcast_to
sparse.solve
Statistical functions: mean, std, var
sparse.isdtype
sparse.reshape
Some elementwise functions
Manipulation functions: concat, expand_dims, squeeze, flip, roll, stack
arg* functions: argmin, argmax
Sorting functions: sort, argsort

IEEE-754 compliance is hard to maintain with sparse arrays in general. This is now even more true of the Julia backend, which trades off performance for IEEE-754 compatibility.