Add insert_maybe_exist interface to fd_map_dynamic

[palmtenor]

In many occasions we could be in the situation where the key we are trying to insert is already in the map. We often need to do a bit different processing (different response, update timestamp, etc.) for that case, hence currently we are doing a lot of query-check-insert, which would do map lookup twice.

This commits adds a insert_maybe_exist interface the either return the existing map entry or insert a new one, and communication about which case it is to the caller.

[kbowers-jump]

Does fd_map_slot.c upsert already cover this case? ``` // mymap_upsert:
//
// ... join is a current local join and key points to the key of
// ... interest (for maps that use key sentinels, *key should not
// ... be a sentinel key), retains no interest in join or key.
//
// ele = mymap_upsert( join, key );
//
// if( FD_UNLIKELY( !ele ) ) {
//
// ... element *key is not in map and the map is too full to
// ... insert it
//
// } else if( FD_UNLIKELY( mymap_ele_is_free( ele ) ) ) {
//
// ... element *key was not in map and should be inserted at
// ... ele. See insert for details how to complete inserting
// ... ele.
//
// ... Note that if mymap uses key sentinels to indicate when
// ... an element is free, this code path will never be taken.
// ... That is, the caller will not be able to tell via
// ... mymap_ele_is_free if an upsert is inserting or updating
// ... as the upsert initialized ele's key field (implicitly
// ... marking the element as not free such).
//
// } else {
//
// ... element *key is stored in the map at ele. See update
// ... for details how to complete updating ele.
//
// ... Note that if mymap uses key sentinels to indicate when
// ... an element is free, we might actually be inserting ele
// ... here (see note above).
//
// }

 myele_t * mymap_upsert( mymap_t * join, ulong const * key );```

[palmtenor]

I think mostly yes. But mymap_upsert is not available in the non-slot (fd_map_dynamic) map?
fd_map_dynamic also doesn't have the explicit entry-free-indicator (like mymap_ele_is_free) to make the upsert result more clear... (if it's an insert or update)

[palmtenor]

Actually I see your point now, we didn't have fd_map_slot in our FD commit...