Fix: missed culture in date-time conversions

[Shiko1st]

Changes fix problems with missed culture in parsers and tests failed when local culture is non-western.

1. Both parsers didn't use configured culture in default excel date conversion.
2. All date-time converters now use the same function to convert incoming string to double. The function tries to adjust the incoming string into form with invariant decimal separator and then parse it to double using invariant number format.
3. Both ExcelDateTimeOffsetConverterTests and ExcelDateTimeOffsetTextConverterTests had hardcoded timeshift in test data. This was rewritten with respect to actual local time shift.
4. Parser tests were generalized, hardcoded strings and values were eliminated. The same data as used when writing sample sheet is used to compare with. The same culture as configured for parser is now used in comparison to convert date-time values to strings.
5. Writer tests were generalized, hardcoded strings and values were eliminated. The same data as used when writing is used to compare with. Writing map is taken into account when performing validation. We belive that converter works correctly as it tested elsewhere so we are free to use it to convert stored value back instead of manually convert it each time. The very big test file was splitted into several parts (with mock data, with general purpose assertion methods, with test methods, with class maps).

The planned development is to use non-US and non-Invariant cultures to test parsers and writer.
p.s. sorry for such a big rewrite but I'm very interested in this library and its evolution so I was needed to ensure it will work and pass tests in my environments. Thas is why I've started all this.