Add new options for convert to allow/disallow accepting different input and invalids with replacement strings

[ScottPJones]

Currently, there is a single convert method that allows conversion to UTF8String with a argument for a replacement string, but that is only defined for an input of Vector{UInt8}, and not from the very many different possible arguments to convert(UTF8String, data).
There are other places in the code where a fixed replacement, 0xfffd, is hard-coded.

1. There needs to be consistent handling for converting strings from vectors of code units (UInt8, UInt16, and UInt32) to the (current) 4 types of strings ASCIIString,UTF8String,UTF16String,UTF32String, that allows specifying the handling of Modified UTF-8, CESU-8, overly long encodings, and also whether invalid sequences should cause an exception, be replaced by a default replacement character (0xfffd for Unicode, 0x1a [SUB] for ASCII), or replaced by a user supplied string.
2. This also needs to work with fixes for convert function for UTF-16 from an AbstractArray{UInt8} problems #11501 and Problems with convert from AbstractArray{UInt8} to UTF32String #11502
3. As laid out in RFC: Roadmap for improving string support in Julia #11558, make sure only valid strings are produced (for security reasons and in the future performance reasons)

[KristofferC]

Moved to JuliaStrings/LegacyStrings.jl#12.