Fix week number from Datetime::week() when weekstart=0

[smemsh]

When working on GothenburgBitFactory/taskwarrior#3623, after modifying the code to pass the user-configured rc.weekstart down into Datetime::weekstart, it was discovered that the returned week number from Datetime::week() was one more than it should be when weekstart=0 (but not weekstart=1). The code was doing:

char weekStr[3];
struct tm* t = localtime (&_date);
strftime (weekStr, sizeof (weekStr), "%U", t);
int weekNumber = strtol (weekStr, NULL, 10);
if (weekstart == 0)
  weekNumber += 1; // <-- HERE

This increment was never activated before, because Datetime::weekstart was initialized to 1 in the library, and never altered by TW. Now that it can be set using configuration (in forthcoming patch), there was breakage seen with weekstart=0 (but not weekstart=1). This was unearthed by TW tests:

not ok 20 - dom2.test.py: DOM 3.due.week
# FAIL: AssertionError [...]
#       '35
# ' != '36
# '
#       - 35
#       + 36

I followed it in a debugger to the [apparently bogus] increment in Datetime::week(). Example that breaks without the patch:

task add testing
task mod 1 due:2011-09-01
task rc.weekstart=monday _get 1.due.week # --> 35 (correct)
task rc.weekstart=sunday _get 1.due.week # --> 36 (incorrect)

The week of 2011-09-01 should be 35 regardless of weekstart:

 $ env - ncal -w -M 09 2011
    September 2011
Mo     5 12 19 26
Tu     6 13 20 27
We     7 14 21 28
Th  1  8 15 22 29
Fr  2  9 16 23 30
Sa  3 10 17 24
Su  4 11 18 25
   35 36 37 38 39

 $ env - ncal -w -S 09 2011
    September 2011
Su     4 11 18 25
Mo     5 12 19 26
Tu     6 13 20 27
We     7 14 21 28
Th  1  8 15 22 29
Fr  2  9 16 23 30
Sa  3 10 17 24
   35 36 37 38 39

 $ date -d 2011-09-01 +%V
35

 $ date -d 2011-09-01 +%U
35

[djmitche]

Oops, I hadn't realized this wasn't merged!