Improve memory use by setting map capacity in dataframe.Row()

[teastburn]

I noticed this while benchmarking with funcs.Evaluate. Here's my before and after for a 45x45 dataframe of float64s with an expression that uses all 45 columns.

BenchmarkFuncEvaluateOriginal-10     	     9235	    625042 ns/op	  853558 B/op	    7564 allocs/op
BenchmarkFuncEvaluateWithMemoryRes-10    	10000	    567637 ns/op	  754678 B/op	    7401 allocs/op