`INSERT` plans could generate large cache entries

[hatyo]

If an INSERT plan is attempting to add a relatively large number of rows to a table, it might end up generating a relatively large cache entry, having too many of these plans stored in the cache could lead to OOM without an aggressive eviction strategy.

For example, a query like:

insert into t((a, b, c, d, e, f, d), (I, j, k, l, m), n, o, p)
values 
(('x1A', 'x2', 'x3', 'x4, 'x5', 'x6', 'x7'), ('x8', 'x9', 'x10', 'x11', 'x12'), 'x13', 42, [43, 44, 45, 46, 47, 48, 49, 49, 49, 49, 49]),
(('x1B', 'x2', 'x3', 'x4, 'x5', 'x6', 'x7'), ('x8', 'x9', 'x10', 'x11', 'x12'), 'x13', 42, [43, 44, 45, 46, 47, 48, 49, 49, 49, 49, 49]),
(('x1C', 'x2', 'x3', 'x4, 'x5', 'x6', 'x7'), ('x8', 'x9', 'x10', 'x11', 'x12'), 'x13', 42, [43, 44, 45, 46, 47, 48, 49, 49, 49, 49, 49]),
...
(('x1Z', 'x2', 'x3', 'x4, 'x5', 'x6', 'x7'), ('x8', 'x9', 'x10', 'x11', 'x12'), 'x13', 42, [43, 44, 45, 46, 47, 48, 49, 49, 49, 49, 49])

ends up generating hundreds of RCVs containing hundreds of Type instances, moreover, the generated QueryPlanConstraint will end up with a boolean expression (QueryPredicate) containing thousands of terms. Profiling shows that the overall retain size is several megabytes.