I noticed srfi-19 date-week-number seems to be off by 1 in its week
start calculation.  For instance looking at Jan 1984, Sat 7th should
be in the first week and Sun 8th begin the next, but alas,

	(date-week-number (make-date 0 0 0 0 7 1 1984 0) 0)
	(date-week-number (make-date 0 0 0 0 8 1 1984 0) 0)

both give 1.

The code looks about right, just doesn't take into account
date-year-day starting from 1 rather than 0.

        * srfi-19.scm (date-week-number): Correction, day of week starting
        week was off by one.

        * tests/srfi-19.test (date-week-number): Add tests.

I think this could go in the 1.6 branch too.