Memory optimization of pruning pass

[rdragos]

Closes #967.

[rdragos]

Stats for the pruning pass, i.e. running

elk compile lowering.moose pruning.moose -p prune

Before (ca7a4dd)

calls to allocation functions: 248494957 (1131306/s)
temporary memory allocations: 43746471 (199161/s)
peak heap memory consumption: 5.18G
peak RSS (including heaptrack overhead): 5.47G
total memory leaked: 173.74K

After:

calls to allocation functions: 225867598 (1153698/s)
temporary memory allocations: 45630111 (233071/s)
peak heap memory consumption: 4.05G
peak RSS (including heaptrack overhead): 3.55G
total memory leaked: 173.74K

In release mode:

before (ca7a4dd)

/usr/bin/time ./target/release/elk compile lowering.moose pruning.moose -p prune 
17.50user 4.95system 0:11.72elapsed 191%CPU
 /usr/bin/time ./target/release/elk compile lowering.moose pruning.moose -p prune
17.89user 4.96system 0:12.07elapsed 189%CPU
/usr/bin/time ./target/release/elk compile lowering.moose pruning.moose -p prune
17.69user 4.87system 0:11.80elapsed 191%CPU

after:

/usr/bin/time ./target/release/elk compile lowering.moose pruning.moose -p prune
15.68user 4.38system 0:09.49elapsed 211%CPU
/usr/bin/time ./target/release/elk compile lowering.moose pruning.moose -p prune
15.06user 4.55system 0:09.07elapsed 216%CPU
/usr/bin/time ./target/release/elk compile lowering.moose pruning.moose -p prune
15.34user 4.59system 0:09.38elapsed 212%CPU

[mortendahl]

🚀

[mortendahl]

nit: add comment that this works before retain is promising to visit elements in their original order?