Add elemwise ops for GCXS

[daletovar]

• Moves element-wise operations
• For now, just converts GCXS to COO to compute elemwise and then converts again at the end
• If COO and GCXS interact, the output will be COO
• Moves methods that rely on elemwise to _sparse_array: round, clip, astype, mean, var, std
  In the future I like the idea of converting all SparseArrays to a linearized representation and do the computations on that and then convert the result to the desired format. It looks like this is basically what's done anyway considering that _match_arrays works with linearized coords. I just think it's currently wasteful to convert GCXS -> COO -> linearized -> COO -> GCXS.

[codecov]

Codecov Report

Merging #412 into master will increase coverage by 0.10%.
The diff coverage is 96.62%.

@@            Coverage Diff             @@
##           master     #412      +/-   ##
==========================================
+ Coverage   95.06%   95.17%   +0.10%     
==========================================
  Files          19       19              
  Lines        2819     2818       -1     
==========================================
+ Hits         2680     2682       +2     
+ Misses        139      136       -3     

[hameerabbasi]

In the future I like the idea of converting all SparseArrays to a linearized representation and do the computations on that and then convert the result to the desired format.

Unfortunately linearized coords don't work for broadcasting cases.

[hameerabbasi]

Thanks, @daletovar. Overall, this looks pretty awesome. I've suggested a couple of minor changes.

[hameerabbasi]

Suggested change

	self.data = other.data
	self.indices = other.indices
	self.indptr = other.indptr
	self.compressed_axes = other.compressed_axes
	super().__init__(other.shape, fill_value=other.fill_value)
	return self.copy(deep=False)

[daletovar]

I don't quite see how this makes a copy of other. I tried this and it makes most of the tests fail.

[hameerabbasi]

This was a misunderstanding on my part.

[daletovar]

Gotcha, no worries.

[hameerabbasi]

Could you try self.__dict__ = other.__dict__.copy()?

[daletovar]

Yeah, definitely. Should I change it for COO as well?

[daletovar]

Thanks so much for the review, @hameerabbasi.