In test/engine/test_transaction/test_two_phase_recover(), a COMMIT
PREPARED is issued while in a transaction. This causes an error, and
a prepared transaction is left hanging around which causes
the subsequent test to hang. I've altered the test to execute the
offending query with autocommit=true, then when it gets to the COMMIT
PRPARED it can go ahead.

There's another complication for pg8000 because its tpc_recover() method
started a transaction if one wasn't already in progress. I've decided
that this is incorrect behaviour and so from pg8000-1.9.13 this method
never starts or stops a transaction.

The pg8000 dialect checks the text of the exception to determine if the
connection is closed. I'd (recklessly!) changed the text of the
exception in a recent version of the pg8000 driver adding capitalization
and a full stop. I've changed it back now so all works.

simplify tox again now that we can exclude tests more easily

rules, better message formatting

- add support for IDENTITY INSERT setting for INSERT with inline VALUES

where engine-level isolation level parameter would raise an error
on connect. fixes #3134

pg8000 uses binary transfer of floats and so accuracy is lost for floats
that can't be represented exactly in binary.

We can now update the requirements to remove the special case of pg8000
truncating decimals.

it for DELETE would fail to target the correct row for DELETE.
Then to compound matters, basic "number of rows matched" checks were
not being performed.  Both issues are fixed, however note that the
"rows matched" check requires so-called "sane multi-row count"
functionality; the DBAPI's executemany() method must count up the
rows matched by individual statements and SQLAlchemy's dialect must
mark this feature as supported, currently applies to some mysql dialects,
psycopg2, sqlite only. fixes #3006
- Enabled "sane multi-row count" checking for the psycopg2 DBAPI, as
this seems to be supported as of psycopg2 2.0.9.