Separate GUC for replication origins

Hi,

We usually use max_replication_slots as the maximum number of replication
origins. It predates the logical replication infrastructure (commit
5aa2350426c). However, it is confusing (if it is the first time you are dealing
with logical replication. Why do I need to set replication slots on subscriber
if the replication slots are stored on publisher?) and it has a limitation (you
cannot have a setup where you want to restrict the number of replication slots
and have a high number of subscriptions or vice-versa). The initial commit also
mentions that it is not adequate (origin.c).

/*
* XXX: max_replication_slots is arguably the wrong thing to use, as here
* we keep the replay state of *remote* transactions. But for now it seems
* sufficient to reuse it, rather than introduce a separate GUC.
*/

and another comment suggests that we should have a separate GUC for it.

/*                                                                              
* Base address into a shared memory array of replication states of size        
* max_replication_slots.                                                       
*                                                                              
* XXX: Should we use a separate variable to size this rather than              
* max_replication_slots?                                                       
*/

The commit a8500750ca0 is an attempt to clarify that the max_replication_slots
can have different meanings depending on the node role (publisher or
subscriber).

I'm attaching a patch that adds max_replication_origins. It basically replaces
all of the points that refers to max_replication_slots on the subscriber. It
uses the same default value as max_replication_slots (10). I did nothing to
keep the backward compatibility. I mean has a special value (like -1) that
means same value as max_replication_slots. Is it worth it? (If not, it should
be mentioned in the commit message.)


--
Euler Taveira
EDB   https://www.enterprisedb.com/