EDR garbage collection in multi-chain cases

[kanej]

In Hardhat 3 we can create multiple EDR instances in one execution. We want to ensure we are correctly unallocating resources once an EDR instance is finished with.

We should investigate that EDR instances are being garbage collected when expected and that we free up the resources on the EDR side when this happens. In particular there is a danger that Hook Handlers hold on to instances of EDR if not carefully implemented.

This is a risk now, but also with changes in Hooks in the future. We should implement a harness and regression test that allows us to detect memory leaks around EDR instances.