interpreter really slow when dealing with large arrays

[proppy]

When working on https://github.com/google/xls/pull/454, we noticed that the interpreter can be very slow when dealing with large arrays.

See the following execution times for bazel run -c opt //xls/dslx:interpreter_main -- $(readlink -f xls/examples/hack.x) depending on the ram and rom array.

rom size	ram size	real time
32	32	0m0.501s
256	256	0m1.380s
512	512	0m4.022s
1024	1024	0m21.773s
4096	4096	9m45.161s

[proppy]

Please find attached the verbose logs for ram and rom size 1024: hack.1024.x.log