pg: ensure all tables have full replica identity

[jchappelow]

resolves #1025

[jchappelow]

This is working now.

Starting with a kwildb-snapshot.sql.gz from v0.8, start with v0.9 in genesis_state using that file (genesis apphash "z0YcmAr5MDZHpgdRyWH4QAiKzMt0yZwb8lWEFr8Ji4A="):

v0.9 main

# \d+ ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.users
                  Table "ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.users"
  Column  |  Type  | Collation | Nullable |   Default   | Storage  | Compression | Stats target | Description 
----------+--------+-----------+----------+-------------+----------+-------------+--------------+-------------
 id       | bigint |           | not null |             | plain    |             |              | 
 username | text   |           |          | 'sds'::text | extended |             |              | 
 age      | bigint |           |          |             | plain    |             |              | 
 wallet   | text   |           |          |             | extended |             |              | 
Indexes:
    "users_pkey" PRIMARY KEY, btree (id)
    "users_wallet_key" UNIQUE CONSTRAINT, btree (wallet)
Check constraints:
    "users_age_check" CHECK (age >= 0)
Referenced by:
    TABLE "ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.posts" CONSTRAINT "posts_user_id_fkey" FOREIGN KEY (user_id) REFERENCES ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.users(id) ON UPDATE CASCADE ON DELETE CASCADE
Publications:
    "kwild_repl"
Access method: heap

v0.9 this PR

# \d+ ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.users
                  Table "ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.users"
  Column  |  Type  | Collation | Nullable |   Default   | Storage  | Compression | Stats target | Description 
----------+--------+-----------+----------+-------------+----------+-------------+--------------+-------------
 id       | bigint |           | not null |             | plain    |             |              | 
 username | text   |           |          | 'sds'::text | extended |             |              | 
 age      | bigint |           |          |             | plain    |             |              | 
 wallet   | text   |           |          |             | extended |             |              | 
Indexes:
    "users_pkey" PRIMARY KEY, btree (id)
    "users_wallet_key" UNIQUE CONSTRAINT, btree (wallet)
Check constraints:
    "users_age_check" CHECK (age >= 0)
Referenced by:
    TABLE "ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.posts" CONSTRAINT "posts_user_id_fkey" FOREIGN KEY (user_id) REFERENCES ds_x708dd55e1dda0fdcd727dc1813d941184622eb1223b1d744e8a49a2e.users(id) ON UPDATE CASCADE ON DELETE CASCADE
Publications:
    "kwild_repl"
Replica Identity: FULL
Access method: heap

It now has Replica Identity: FULL.

[jchappelow]

When unset it defaults to what we were setting anyway. I've deprecated this field since there are numerous assumptions about "ds_", and if we set this field to any other function nothing would work.

[brennanjl]

I dont think we should change it now, but is the "ds_" necessary? The original reason we prepended DBIDs with the "x" is so that they could be used as Postgres schemas (since Postgres schemas must start with a letter).

The only reason I bring it up is b/c some people (like idOS) have begun modifying their schemas with Postgres directly while doing migrations. If this becomes more common user behavior, it might be easier to just make schema names 1-1 with the dbid.

Not anything we need to address right now, just food for though.