Remove some Float64 assumptions

[ericphanson]

Fixes #876.

I also changed convert(T, √2) to √T(2) in a few places to improve precision when T is a BigFloat (calculate the square-root in the BigFloat domain).

[ericphanson]

Note: I didn't add any tests, because I think the tests solution is to finish #841 and then test those tests with BigFloat problems. But I can say from local testing that these changes let SDPA-GMP pass Convex.jl's tests with the SOCP problems that otherwise fail.

[codecov-io]

Codecov Report

Merging #878 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #878      +/-   ##
==========================================
+ Coverage   95.01%   95.01%   +<.01%     
==========================================
  Files          80       80              
  Lines        8284     8286       +2     
==========================================
+ Hits         7871     7873       +2     
  Misses        413      413

Impacted Files	Coverage Δ
src/Bridges/Constraint/rsoc.jl	100% <100%> (ø)	⬆️
src/Bridges/Variable/soc_to_rsoc.jl	100% <100%> (ø)	⬆️
src/Bridges/Constraint/slack.jl	100% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 345c03b...4e6275f. Read the comment docs.