add option to clean zeros on SDP cuts after form outer products

[chriscoey]

hopefully this will fix the failures and numerical instability warnings we get on SDP tests

maybe a new tolerance, 0 for no cut cleaning, some small positive number for cut cleaning

[chriscoey]

this did not fix the failure of:

SDP+SOC conic MISOCP - Iter, Pajarito, SCS  |   82     4     86
    Init SOC cuts: JuMP.jl A-opt design         |          4      4

which was Pajarito(CPLEX, Mosek) as the MISOCP solver

[chriscoey]

until we have stronger evidence that cleaning zeros on SDP cuts is needed, this can be closed