Remove `:noshift` SubString constructor and replace it with something else

[jakobnissen]

In commit 7489a31, a new constructor for SubString was added:

function SubString{T}(s::T, i::Int, j::Int, ::Val{:noshift}) where T<:AbstractString
    @boundscheck if !(i == j == 0)
        si, sj = i + 1, prevind(s, j + i + 1)
        @inbounds isvalid(s, si) || string_index_err(s, si)
        @inbounds isvalid(s, sj) || string_index_err(s, sj)
    end
    new(s, i, j)
end

This is undocumented, but it's a method of the public SubString, and also tested.
I 100% sympathize with the need for having a substring constructor without checks that compiles to a noop.

However, I don't think it's a good method:

• It makes it way too easy to accidentally create a SubString{SubString{T}}, which I don't think ought be encouraged (or allowed, but that ship has sailed)
• It's not clear from the signature that i and j refers to offset and length, not start and stop indices, as it does in the other methods. This is very confusing.
• It's a little strange with the Val{:noshift} as an argument.
• Another minor issue is that the bounds- and index checks causes the constructor to not inline, even when annotated with @inbounds. So, it needs to be used @inbounds @inline ..., which nobody is going to remember

It might possibly be breaking to remove this method. It's undocumented, but a method of a public type.
The only usage in the wild according to JuliaHub code search is JuliaSyntaxHighlighting.jl.

The method was added in #49586 without mentioning it, possibly it was meant to be a private method. Cc. @tecosaur.

I suggest creating a new (public or exported) function which allows users to skip index and bounds computations. Possibly unsafe_substring(s::AbstractString, first_codeunit::Int, last_codeunit::Int).

[KristofferC]

Ref #51923 (comment).

[LilithHafner]

Because this is not documented, it is not part of the public API, so this is not a minor change, simply a change to internals.

[LilithHafner]

Triage agrees with Lilith.