Remove unused code in parsing and speed parsing up

[anowacki]

The NanoDate(::AbstractString, ::DateFormat) constructor previously
created a few variables which are never used, so remove these lines.
These lines are actually responsible for about half the running time of
calls to the constructor, so this change doubles the speed of parsing.

Also remove the unused separate_offset methods.

[anowacki]

Before (v1.0.2):

julia> @benchmark NanoDate("1234-01-02T01:23:45.123456789+01:23")
BenchmarkTools.Trial: 10000 samples with 3 evaluations.
 Range (min … max):  8.617 μs … 28.051 μs  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     9.091 μs              ┊ GC (median):    0.00%
 Time  (mean ± σ):   9.873 μs ±  1.543 μs  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▁▆██▇▆▄▂      ▁▃▃▃▃▂▁ ▁▁▁▂▂▃▃▃▂▂▁▁▁▁▁▂▁▁▁                  ▂
  █████████▇▇▇█████████████████████████████████▆▇█▇▇▇▇▇▇▆▆▅▅ █
  8.62 μs      Histogram: log(frequency) by time     14.9 μs <

 Memory estimate: 2.66 KiB, allocs estimate: 52.

After (016ccae):

julia> @benchmark NanoDate("1234-01-02T01:23:45.123456789+01:23")
BenchmarkTools.Trial: 10000 samples with 7 evaluations.
 Range (min … max):  5.056 μs …  10.450 μs  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     5.241 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   5.297 μs ± 325.704 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

    ▃██▅                                                       
  ▃▆█████▅▃▃▂▂▂▂▂▂▂▂▂▂▂▂▂▂▂▂▂▂▂▁▂▂▁▁▁▁▂▂▂▁▂▂▁▁▁▁▁▁▁▁▁▁▂▂▂▂▂▂▂ ▃
  5.06 μs         Histogram: frequency by time        7.51 μs <

 Memory estimate: 1.88 KiB, allocs estimate: 43.

[JeffreySarnoff]

Much appreciated.