RFR: JDK-8317772: NMT: Make peak values available in release builds

[Thomas Stuefe]

We added peak counts to NMT with [JDK-8297958](https://bugs.openjdk.org/browse/JDK-8297958). That feature is *extremely* helpful in analyzing native memory footprints since it gives you a glance into the past. For instance, it makes libc memory retention easier to explain.

Initially, I restricted this feature to debug JVMs out of vague concerns about performance when running with NMT. However, after measuring, it turns out that the increase in cost is very small. This makes sense since we only update the peak values when we actually reach a peak. 

Since the payoff is huge, it makes sense to have this feature always enabled. We don't need a new switch for that. As a side effect, NMT code becomes a bit clearer since we can lose a number of assert ifdefs.

-------------

Commit messages:
 - JDK-8317772-NMT-Make-peak-values-available-in-release-builds

Changes: https://git.openjdk.org/jdk/pull/16119/files
 Webrev: https://webrevs.openjdk.org/?repo=jdk&pr=16119&range=00
  Issue: https://bugs.openjdk.org/browse/JDK-8317772
  Stats: 13 lines in 3 files changed: 0 ins; 7 del; 6 mod
  Patch: https://git.openjdk.org/jdk/pull/16119.diff
  Fetch: git fetch https://git.openjdk.org/jdk.git pull/16119/head:pull/16119

PR: https://git.openjdk.org/jdk/pull/16119