Redundant check in recover_polynomial_coeff?

[kevaundray]

This has been taken from Benedikt Wagner's document.

You already check that:
-  (1) cell indices contain no duplicates
-  (2) all cell indices are between 0 and CELLS_PER_EXT_BLOB. 
This should imply that there at most CELLS_PER_EXT_BLOB many cell indices.

I believe this code should be spec compliant, so this might imply a spec change is needed too.

Related to #246

[b-wagn]

Note: this comment was about this part of the code.

[b-wagn]

Also, you are right: this is also explicitly checked in the spec:

assert CELLS_PER_EXT_BLOB / 2 <= len(cell_indices) <= CELLS_PER_EXT_BLOB in this function.

To be spec cimpliant, we should have it explicitly. This also has the advantage of clearly stating the assumptions we make.
Feel free to just close the issue, or leave a comment saying it may be redundant in the code.