Solidly curves

[sklbancor]

This is the second part of the Solidly curve implementation (part one is here). It adds a few further analytics, but more importantly, it now adds the ability to add Solidity curves to the optimizer.

This is implemented via the new CPC method from_solidly that should be called with the two token balances x and y (obviously consistent with the pair provided, in tokens as floats).

VERY IMPORTANTLY, THE METHOD RETURNS A LIST OF CURVERS, RATHER THAN A SINGLE CURVE. The calling code should be able to handle the following situations

1. Exactly one curve is contained in the list; this is the current default case when the match works, ie the AMM is within the linear area and not in the wings
2. The list is empty, which means that the AMM is in the wings, and our flat approximation does not work, so we ignore it
3. There are multiple curves contained in the list; this case is not implemented yet, but it may be implemented soon. In this case, the traded against the Solidity curve should the the sum of all trades against the returned curves_

Integration instructions

The code in this PR is now ready for integration:

• get the (x,y,pair) values from the relevant solidly curves
• call CPC.from_solidly(...)
• this returns a list and the following cases should be dealt with (even though the last can't at the moment happen, but it may be easier to deal with it right away rather than having to change the code later again)
  • empty list -> the Solidly AMM is outside the range we approximate and should be ignored
  • one element -> the Solidly curve is approximated by a single CPC curve
  • multiple elements -> the Solidly curve is appriximated by multiple CPC curves
• when creating trade instruction, all dx and dy values across all curves returned should simply be added up* to obtain our best estimate of how the curve will react; note that this is an approximation and it may be off by up to one percent at the current calibration and the code may have to deal with this

*as discussed previously, currently we only have one curve; when we have multiple curves you can be sure that the signs of the dx will all be the same (or zero), and ditto for the dy; most likely all but one will be maxed out, but you should not rely on it (and I don't think you care anyway)