Add zipShortest function and make generic Zip safe

[3Rafal]

Added safe zip function zipShortest.
It's behavior is equivalent to Haskell zip which I think is quite reasonable.
Naming is already used in .NET ecosystem in MoreLinq C# library [Source]

[adz]

I like your idea!

I thought this was how List.zip in core worked already, but I was wrong. It throws exception if different sized lists are zipped.

Unfortunately, F#+ seems a little inconsistent:

• extensions for Enumerator, Dictionary, IReadOnlyDictionary, Map implement zip just like your zipShortest (i.e. it stops when one enumerator stops, and skips over missing keys for map, etc).
• NonEmptyList.fs and NonEmptySeq.fs in FSharpPlus/Data directly call List.zip / Seq.zip, so these throw.

So for completeness, we should probably implement zipShortest for NonEmptyList, NonEmptySeq.

I'm not sure if it's worth also implementing zipShortest as zip in the other cases?

[adz]

Oh now I notice the other pr!

Well seems to me still makes sense to implement for non empty list... tryZip could be a separate thing.

[3Rafal]

@adz , you're right. I've added implementation for nelist and neseq as you suggested.
When it comes to tryZip, I'm not convinced it is a good idea, so I won't implement it now.

[gusty]

Unfortunately, F#+ seems a little inconsistent:

Again it worth clarifying is not our goal to fix FSharp.Core
If they took the unfortunate decision to make lists and arrays to fail when zipping we have to honor it, if we have a nelist it can't behave differently to the normal list, but same goes for neseq.

So these inconsistencies comes from F# core and we can't fix them. We should instead create different function which behave in a consistent way, take for instance the example of take and skip, they fail but we introduced other functions drop and limit that don't.

[gusty]

We could however, define our generic zip as zipShortest, that would solve inconsistencies at the generic level.
Same would apply for map2.

But I definitely like the zipShortest in the extensions.

[3Rafal]

@gusty , I've introduced changes according to your remarks. now generic zip is using safe functions only.

[gusty]

I will accept it as it is, but not sure if this is the most efficient implementation in terms of allocations.

[3Rafal]

Changed it to efficient version both here and in ResizeArray.
I've checked it with #time and implementations suggested by you are considerably faster. (even 2-3x on my machine)