sql/analyzer/indexes.go: Avoid N^2 behavior in index merge for large indexed InTuple expressions.

[reltuk]

No description provided.

[reltuk]

This improves the behavior reported here: dolthub/dolt#1511. For example:

$ time dolt sql < query_2000.sql
+----+-------+
| pk | value |
+----+-------+
| 1  | 1     |
| 2  | 2     |
+----+-------+
Rows inserted: 0 Rows updated: 0 Rows deleted: 0
dolt sql < query_2000.sql  0.10s user 0.04s system 67% cpu 0.213 total

There are still potential improvements here. For large result sets, we still evaluate the n entries in the tuple for each row when evaluating the expression as part of a projection or filter expression, for example.

[zachmu]

LGTM!