Prior to this fix, performance schema configuration variables that
can be automatically sized by the server were given incorrect values
in some cases.

Performance overhead can increase significantly
with some pathological cases created by incorrect sizing,
causing this bug to have a real impact.

For example, for 50,000 tables in the table open cache,
the performance schema estimated that a value of 20,000 table handles
was adequate. This is wrong, a proper value should be 100,000 in this case.

The problem is that the heuristic used values that got altered by 
a 'sanitizing' step, supposed to prevent bad usages.

The fix is to simply remove this sanitizing step, because:
- it does not resolve any problem by itself, only creates more,
- it introduces hard coded values which are not based on real deployments in
  production

With this fix, heuristics computed by the performance schema for automated 
sizing are expected to be more adequate, for high end configurations.