Re: Review: Row-level Locks & SERIALIZABLE transactions, postgres vs. Oracle

This patch suffered a lot of bitrot in the last month, mostly due to the
noop patch for SSI. It also used the term SERIALIZABLE in many places to
indicate any transaction-snapshot mode transaction, so I applied changes
consistent with the noop patch. I also found a few whitespace and brace
usage issues which seemed to conflict with conventional usage and cleaned
those up. A minor spelling typo also corrected.

A new patch reflecting all this is attached.

This now compiles and passes regression tests. I still need to re-run all
the other tests which Florian and I previously used to test the patch. I
don't have any reason to expect that they will now fail, but one need to be
thorough. Once that is confirmed, I think this will be ready for committer
unless someone can think of something else to throw at it first.

-Kevin