Re: pgsql: Improve performance of timezone loading, especially pg_timezone_

I wrote:
> I concur with your thought that these are separate bugs and deserve
> separate commits. Will do that in a minute.

BTW, I wrote in the commit log message that the posixrules omission
had only minimal consequences, but on closer look that's wrong:
the hardwired default in the timezone code is

/*
* The DST rules to use if TZ has no rules and we can't load TZDEFRULES.
* We default to US rules as of 1999-08-17.
* POSIX 1003.1 section 8.1.1 says that the default DST rules are
* implementation dependent; for historical reasons, US rules are a
* common default.
*/
#define TZDEFRULESTRING ",M4.1.0,M10.5.0"

that is, first Sunday of April and last Sunday of October. The US
hasn't gone by that since 2006. So actually, this change will have
a very visible impact on the behavior of POSIX-style zone names.

(I wonder why IANA hasn't updated this default? Still, that's their
expertise not ours, so I'm disinclined to touch it.)

regards, tom lane