Non deterministic compile results with lamda numbering

In PDE we are experiencing a build instability with the class [BundleLauncherHelper](https://github.com/eclipse-pde/eclipse.pde/blob/master/ui/org.eclipse.pde.launching/src/org/eclipse/pde/internal/launching/launcher/BundleLauncherHelper.java) according to the logs the numbering of lamdas is different compared to those on the baseline, I have captured a compile log that shows the differences and **has compile logs with exact classpath and compiler options** here: [compare.zip](https://github.com/eclipse-jdt/eclipse.jdt.core/files/14088236/compare.zip)

The textcompare shows that only the numbering is different here, please note that the baseline seem not contain any `lambda$0` but starting with `lambda$1` immediately:

I can't reproduce this locally and even on the CI instance the behavior varies between builds where none are actually changing that bundle.

Start with ui/org.eclipse.pde.launching/src/org/eclipse/pde/internal/launching/launcher/BundleLauncherHelper.java and inspect compare.zip, which contains the differing compile logs, classpaths, and compiler options. Compare the lambda numbering between builds and the baseline; done means the cause of the varying numbering is identified and compilation is deterministic.

Written by the indexing model from the issue text.