Allow JVM to optimize code before running experiments

[rmartinsanta]

Is your feature request related to a problem? Please describe.
When running a small number of iterations, specially if runtimes are short, the jump between interpret mode and compiled mode by the JVM after executing a method several times is pretty noticeable in the results, and could lead to wrong assumptions by the user.

Describe the solution you'd like
Allow the user to specify a fast instance that will be run before the main experiment for all configurations. This way, most code should be already compiled and cached by the JVM by the time the main experiment run. Additionally, during instance validation, we could guess which instance is the smallest one, if not specified by the user, by smallest load time / smallest file size, requiring zero configuration.

Additional context
https://www.baeldung.com/jvm-tiered-compilation

[rmartinsanta]

The feature can be tested using -XX:+PrintCompilation, the idea is that before the main experiments run, critical code paths should be already compiled and optimized by C2.