Introduce basic Flow<T>.chunked operator

[qwwdfsad]

Fixes #1290

[qwwdfsad]

A few things about this PR:

• It's a long overdue we introduce chunked operator, and whatever we do, chunked(n) is likely to be here as the shorthand syntax
• It also might bootstrap the discussion and break our analysis paralysis
• transform version (chunked(n, transform)) is not added even though the stdlib has one -- we have a strong preference for proper separation of concerns. The documentation sample provides a clear substitute
• Meanwhile, I'll experiment with duration overload, but it's less straightforward

API shape brain dump (eventually I'll move it towards a better-structured issue, but not there yet):

There exists a whole bunch of possibilities, use cases and things full-blown chunked family might/should address

• Kotlin's windowed -- the same, but buffers are overlapped
• Reactive windowed -- the same as chunked, but it yields Flow<Flow<T>>. Operator to chunk a single Flow<T> into a Flow<Flow<T>> #2989
• Time-based chunking:
  • The most straightforward and common -- keep the buffer open for the given duration, then emit it if it's not empty. Start the new buffer
  • A whole variety of supplemental things -- with time skips, predicate-based buffering
• Batching: group while you can (or the limit exceeded), emit it and wait the next element.

The problem with time is that it requires the actual concurrency, while the size-based version does not. I'm still exploring the possibility of something like chunkedTransform { /* decide whether it's time-based or not */ }, but not much of ideas for now.

[dovchinnikov]

Please name the parameter maxSize then

[qwwdfsad]

That was my original intent, but stdlib disagrees on that: https://kotlinlang.org/api/latest/jvm/stdlib/kotlin.collections/chunked.html

A nice thing to consider tho