avoid expensive runtime div in parse for common bases

[KristofferC]

10 and 16 are common bases to parse to so let's avoid the "killer div" in those cases.

Benchmarks:

julia> @btime parse(Int64, "12");
  45.326 ns (0 allocations: 0 bytes)

julia> @btime parse(Int64, "12345678912345");
  84.381 ns (0 allocations: 0 bytes)

julia> @btime parse(UInt8, "0x10");
  47.826 ns (0 allocations: 0 bytes)

julia> @btime parse(UInt64, "0x123aca1231a");
  78.174 ns (0 allocations: 0 bytes)

After

julia> @btime parse(Int64, "12");
  36.399 ns (0 allocations: 0 bytes)

julia> @btime parse(Int64, "12345678912345");
  74.086 ns (0 allocations: 0 bytes)

julia> @btime parse(UInt8, "0x10");
  39.586 ns (0 allocations: 0 bytes)

julia> @btime parse(UInt64, "0x123aca1231a");
  72.017 ns (0 allocations: 0 bytes)

I have some thought on how this overflow handling can be done a bit better, but for now, might as well do the incremental thing.

[KristofferC]

@nanosoldier runbenchmarks("parse", vs = ":master")

[JeffBezanson]

Should these be - T(9) and - T(15)?

[KristofferC]

Oops, should have ran the tests first heh.

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan