Investigate 2 errors emitted in the optimizer

[barakman]

Bug Description

There are 2 different errors which are occasionally emitted in the optimizer:
The 1st error:

RuntimeWarning: overflow encountered in exp
p = np.exp(plog10 * np.log(10))

The 2nd error:

RuntimeWarning: invalid value encountered in scalar divide
p = pvec[self.tknx] / pvec[self.tkny]

Neither one of these errors ends up with an uncaught exception, i.e., both are handled gracefully in some way.

The result is a potentially-undesired behavior of the system.

Severity

High

Steps to Reproduce

Both errors are consistently reproducible in test_901_TestMultiTriangleModeSlow.py:

• The 1st error occurs in function test_test_combos
• The 2nd error occurs in function test_test_find_arbitrage_single

PR #546 attempts to handle the 1st error, so it may serve as a starting-point.

The 2nd error appears to be just a matter of the price of one of the two tokens (and sometimes the prices of both of them) being zero, which is pretty much invalid to begin with.
So a special handling of this specific scenario should possibly be applied here, perhaps at a much earlier point, before even entering optimizer's code.
Alternatively, the initial state could actually be a valid one, but with some inaccurate or incorrect calculation along the way ultimately leading to an invalid (zero) price.