improve performance of conv

[ericphanson]

julia> using Convex, BenchmarkTools

julia> x = rand(100_000);

julia> y = Variable(200);

julia> @btime conv($x, $y)

gives:

  1.286 s (56 allocations: 1.04 GiB) # master
  562.229 ms (37 allocations: 1.64 GiB) # PR

So 2x speed, but 1.6x allocations. I think it's a win overall, though not huge. Aimed at #376

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.87%. Comparing base (96904be) to head (4ad7a93).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #634   +/-   ##
=======================================
  Coverage   97.86%   97.87%           
=======================================
  Files          88       88           
  Lines        5115     5127   +12     
=======================================
+ Hits         5006     5018   +12     
  Misses        109      109           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.