Add more FieldX instances for tuples

[worm2fed]

Why this instances is commented?
https://github.com/monadfix/microlens/blob/effacc20f861de1a05a04a5b697dcb8690b80521/microlens/src/Lens/Micro/Internal.hs#L423-L441

In package description there are:

This library is an extract from lens (with no dependencies). It's not a toy lenses library, unsuitable for “real world”, but merely a small one. It is compatible with lens, and should have same performance. It also has better documentation.

With this lines commented we receive version of _1 not really equivalent with same in lens

Is there any reason for it or it was just missed for some reasons?

[neongreen]

They probably increased the compilation time beyond our 4s target, though I'm not sure. I can reenable them if you have a usecase for them.

[worm2fed]

We have our Prelude which uses micro-lens and re-exports some it's functions.
And we have our project which uses lenses (because micro doesn't allow to make this:

content = mempty
  & at "multipart/form-data" ?~ mempty & schema ?~ paramSchema

out of the box) -> so we switched to lens

And we also have type which needs instance for 7-tuple.
For now we have same instance as commented here in Orphans, and it's not a big deal. We ok with it.
But it's strange, that there are solution already, but it can't be used by others.

How do you think, is it really significantly increases compilation time?

[neongreen]

For what it's worth, I think for apps you should use lens or optics anyway, not microlens.

We have 5*4/2 = 10 instances for _1 _2 _3 _4 _5. If I add instances up to _9, it will be 9*8/2 = 36 instances, a significant increase. Currently microlens builds in 6.5s; if I add the instances, it builds in about 6.9–8.2s.

So for now I will avoid adding the instances. If anybody else wants them — please chime in!