feat: add fast point-to-scalar multiplication via ecrecover

[StackOverflowExcept1on]

Resolves #18

Explanation of how it works:

• ecrecover(e, v, r, s) works according to the formula $Q = r^{-1} ( sR - eG )$ from SEC-1 v2 section 4.1.6 "Public Key Recovery Operation"
• the formula has some limitations: $v \in [ 27, 28 ], r \in [ 1, Secp256k1.Q ), s \in [ 1, Secp256k1.Q )$
  Secp256k1.Q = 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFEBAAEDCE6AF48A03BBFD25E8CD0364141
  

  crysol/src/Secp256k1.sol

  Lines 147 to 149 in 462e903

  	/// @dev The order of the group generated via G.
  	uint internal constant Q =
  	0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFEBAAEDCE6AF48A03BBFD25E8CD0364141;

• to calculate $scalar \cdot Point$ need to substitute the following values:
  $e = 0, v = PointYParity + 27, r = PointX, s = r \cdot scalar$
• so the formula will look like this after substitution:
  $Q = r^{-1} sR = r^{-1} \cdot r \cdot scalar \cdot R = \cancel{r^{-1}} \cdot \cancel{r} \cdot scalar \cdot R = scalar \cdot R$
• point $R(x, y)$ has coordinates $(r, y)$, where $y$ is calculated from $x$ (coordinate compression) and $v$ (even/odd $y$ is specified)

Warning

This fast multiplication does not work for all points on the curve! For example, PointX >= Secp256k1.Q will cause ecrecover to return address(0). For more details, see https://ethresear.ch/t/you-can-kinda-abuse-ecrecover-to-do-ecmul-in-secp256k1-today/2384/19.

[pmerkleplant]

The code looks very good! Thanks for taking the time and following the projects coding style <3

I'll need some time to review the math (crysol is a learning project after all :P) and think about reasonability of the functionality.

[StackOverflowExcept1on]

As for math, Vitalik approves xD! https://ethresear.ch/t/you-can-kinda-abuse-ecrecover-to-do-ecmul-in-secp256k1-today/2384
Also in thread there is dude k06a who is the co-founder of the 1inch network and this is basically a rewrite of his code.

[pmerkleplant]

Finally managed to jump back into this, sorry for taking so long.

Note that I renamed to function to mulToAddress() which, imho, carries more semantic meaning. Added a note to the function doc about the performance increase though.

Many thanks for the great PR!