Fixes on Types & Memory Plugin

[H4ad]

I didn't split the commits correctly since I did a couple changes that I wanted to discuss first.

So:

• Some types were incorrect, I tried to fix all of them
• Added export for the memory plugin
• Fixed the memory plugin since it was in the old behavior
• Changed how to specify a plugin (class instead of instance)
  • This is a breaking change compared to the old behavior but this will ensure the memory plugin and many other plugins that rely on state per benchmark works correctly.

[RafaelGSS]

I don't think we should land breaking changes at this point (the change of plugins to not need new anymore is a breaking change). I'd focus this PR solely on plugins/memory.js.

[H4ad]

We are on v0, so I don't think the breaking changes are really impactful.

Also, how do we solve the issue of creating new instances for each benchmark when we can't instantiate it?

The change from not need new, to need new was also a breaking change.

[RafaelGSS]

The difference is that this package is now way more used than it was when I did the breaking change https://npmtrends.com/bench-node - I have personally installed it on some automations that I prefer not to break.

Why do we want to create new instances for each benchmark? It's user's responsibility to create one instance for each suite, and our reponsability to .reset after each benchmark fn.

[RafaelGSS]

Also, it would be great if we split this PR into different PRs. It makes my review easier (I know they are already split into commits).

[H4ad]

our reponsability to .reset after each benchmark fn.

There's no reset. But I added again and kept the instance way of having plugins.

Also, it would be great if we split this PR into different PRs. It makes my review easier (I know they are already split into commits).

I simplified the PR, so should be easier to review now.

[H4ad]

I don't remember to update this part but I thikn it makes sense to update, so should I keep it?