Currently, we run compilation of node-profiling on more or less all PRs:

This is pretty slow and the E2E tests have to wait for this to finish, which delays the whole CI run. We should find a way to reduce this overhead/streamline this, as now our CI run takes almost 2x as long as before 😬

Some ideas:

• Can we somehow cache this much more aggressively, maybe? And only run compilation if something inside of node-profiling changed?
• Can we extract/remove/change the node-profiling E2E job so that not all E2E tests have to wait for node-profiling to be compiled?
• Can we only run this if something in node-profiling itself changes (vs. in any downstream package like core, node, ...)?