Add option stream for computing signature

[anh-tong]

The update in this PR allows signax.signature to support stream option.

signax.signature(path, depth, stream=True) will compute signature for stream data. That is, suppose that we have a stream data $\mathbf{x} = (x_1, x_2, \dots, x_n)$, and $\mathbf{x}_k = (x_1, x_2, \dots, x_k)$. With stream=True, the function will return

$$(Sig(\mathbf{x}_2), Sig(\mathbf{x}_3, \dots, Sig(\mathbf{x}_n))$$

Some minor updates:

• using jaxtyping for type anotation (but not clear how we can anotate the type of list[Array] with specific dimension of tensor algebera yet)
• Remove SignatureCombine (it is better to just call signax.signature_combine)
• Add LogSignatureTransform module (with stream option)

[codecov]

Codecov Report

Merging #33 (c42ab29) into main (caf998f) will increase coverage by 5.54%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #33      +/-   ##
==========================================
+ Coverage   87.31%   92.85%   +5.54%     
==========================================
  Files           7        7              
  Lines         205      224      +19     
==========================================
+ Hits          179      208      +29     
+ Misses         26       16      -10     

Impacted Files	Coverage Δ
src/signax/tensor_ops.py	98.33% <ø> (ø)
src/signax/module.py	100.00% <100.00%> (+30.43%)	⬆️
src/signax/signatures.py	100.00% <100.00%> (+2.00%)	⬆️

... and 1 file with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[phinate]

LGTM! As said in slack, we can do a proper jaxtyping PR later to make sure we're hitting all the right checks with shape-correctness!