Make INDEX/2 more efficient

[EugenGez]

Hi,

I think there is some faster (a little bit) and more readable implementation of SQL-ish INDEX(...).

About 10% performance win:

~/.jq

# better implementation?
def INDEX2(stream; idx_expr):
  reduce stream as $row ({}; .[($row|idx_expr|tostring)] |= $row);
def INDEX2(idx_expr): INDEX2(.[]; idx_expr);

$ ./jq-index.sh
+ jq -n '[ range(1024*2000) | {id: . } ] | INDEX(.id)  | empty'

real    0m21.124s
user    0m0.000s
sys     0m0.015s
+ jq -n '[ range(1024*2000) | {id: . } ] | INDEX2(.id) | empty'

real    0m19.165s
user    0m0.000s
sys     0m0.015s

Thank You for the great job!

Regards

[coveralls]

Coverage remained the same at 84.615% when pulling bdcb753 on EugenGez:sql-ish-operators into 7fd9e86 on stedolan:master.

[itchyny]

This improvement was included by #1845 (76e72a3) on Feb 27, 2019. Thank you.