pgsql: Reduce assumptions about locale's behavior in new regex tests.

Reduce assumptions about locale's behavior in new regex tests.

I was overoptimistic to assume that UTF8-based locales would all
consider U+1500 to be a member of the [[:alpha:]] char class.
Tweak the test cases added by commit 78a843f11 to avoid that
assumption.  We might need to lobotomize them further, but this
should be enough to fix the early buildfarm reports.

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/b66336c4e1af0e8eae520623e4b018251807b0bb

Modified Files
--------------
src/test/modules/test_regex/expected/test_regex_utf8.out | 12 ++++++------
src/test/modules/test_regex/sql/test_regex_utf8.sql      |  6 +++---
2 files changed, 9 insertions(+), 9 deletions(-)