This relates #6

Fixes #170 #171

fix missing encoding parameter in setup.py

Spelling.

rel  #162

Closes: taichino/croniter#162

C901 is normally not checked because max-complexity is not set, but
if the user has max-complexity set in their user-wide .config/flake8,
this will prevent false positives.

when crontab day only one day, but last month not has the day, example crontab expression: 0 6 30 3 *

- Remove _get_last_weekday_of_month and replaced it with
  _get_nth_weekday_of_month() and avoid making multiple calls to the calendar
  module when both '#' and 'L' syntax are used at the same time.
- Updated unittests for get_last_weekday_of_month() to instead simply call
  _get_nth_weekday_of_month()[-1].
- Create unit test for new function.

Note that _get_nth_weekday_of_month() is a good candidate for caching as
repeated calls to calendar add up over time, especially when evaluating
multiple cron expressions concurrently.

- Optimize scenario where only 'L' modifier exists in the wday expression to
  reduce the number of invocations of the calendar module.
- Cleanup and combined parsing logic for specialty wday syntax (d#n/ Ld)
- Add new unit tests to confirm proper exception handling for out-of-range 'L'
  and '#' nth month for the wday expression.
- Add examples of new syntax in README.
- Update changelog to cover new syntax and new exception.

- Added new exception class and detection log for the currently known corner
  case where croniter is unlikely to return the correct results.
- Updated unittests to look for this new exception.

- Removed much of my new implementation logic, and merged it into the existing
  nth_weekday_of_momth data structure.  The difference between 5th Saturday
  '6#5' and last Saturday of the month 'L6', in is simply `{'l'}` vs `{5}`.
- Updated tests cases based on new capabilities (which were decently expanded,
  and several limitations and pain points in the code eliminated)!
- Marked some test cases with 'expectedFailure' because apparently in the
  current implementation mixing literal dow values with nth dow isn't properly
  supported.  At least there's no unittests showing that it should work.   You
  get results, but they appear to be wrong to me.  So I'm considering this to
  be a bug, rather than something my new code broke.

- Expanded several few unit tests to show combinations of dow mixing,
  separately and combined, so that if mixed dow mode becomes supported there
  are a few tests to start from.  See tests `test_lwom_mixup_*()`

- Add new unit tests showing how dow L + other should work
- Rename some unittests for consistency

- Commented out the dow_types checking (L + non-L) detection and explicit
  blocking due to mishandled corner cases.