Options OUTPUT_PLUGIN_* controlling format are confusing (Was: Misleading error message in logical decoding)

Hi all,

As there has been a dump of CATALOG_VERSION_NO on REL9_4_STABLE
recently, I am coming back to the options OUTPUT_PLUGIN_* that are
rather confusing for developers of decoder plugins. In short, when
extracting changes from a replication slot, a decoder plugin is free
to set one option that influences the output that is fetched by a
logical receiver: OUTPUT_PLUGIN_TEXTUAL_OUTPUT or
OUTPUT_PLUGIN_BINARY_OUTPUT. The interesting point about this format
option is that it does not directly influence the changes generated by
an output plugin: its value only has effect on the set of functions
pg_logical_[get|peek]_[binary_|]changes that can be used by a
non-replication connection to get individual changes from a repslot.

So a receiver fetching changes using PQgetCopyData is not really
impacted by this format option. Even better it is even possible to use
a custom option that is part of output_plugin_options to switch the
OUTPUT_PLUGIN_* value (cf option force-binary in
contrib/test_decoding).

My point is: logical decoding is presenting in its infrastructure API
a format option that could be added as a custom option in a decoder
plugin. Isn't that orthogonal? To illustrate this argument here is an
example: we could remove force-binary in test_decoding and replace it
by a option that allows the user to directly choose the output format,
like format=[binary|textual] and do the conversion in the plugin. In
the same manner, the functions pg_logical_[get|peek]_binary_changes
are equivalent to their cousin pg_logical_[get|peek]_changes casted to
bytea.

I am raising this point before the 9.4 ship sails, thinking long-term
and to faciliate the maintenance of existing code. Attached is a patch
that simplifies the current logical decoding API regarding that.
Regards,
--
Michael