[WIP] Implement Encoding in rust

[b-n]

By extracting out Encoding into it's own module, we can start to rely on some of the implementations in spinoso-string to provide the functionality.

Some rough todos:

• Implement instance methods
• Implement constants and their binding to the interpreter

Relates: PR #2578
Relates: #183

[lopopolo]

some early thoughts

[lopopolo]

this shouldn't need to be made pub. all of the types are pub use'd below

[lopopolo]

regex is really heavy here. We should be able to use ByteSlice::replace from bstr: https://docs.rs/bstr/latest/bstr/trait.ByteSlice.html#method.replace

[b-n]

Ah nice, that is indeed much better.

I was thinking of whether there is another way to create bare symbols. e.g.

irb(main):001:0> constants = Encoding.constants
=> 
[:ISO8859_14,                                                         
...                                                                   
irb(main):002:0> Encoding::UTF_8.replicate("Super Special Encoding")
=> #<Encoding:Super Special Encoding>
irb(main):003:0> Encoding.constants - constants
=> [:Super_Special_Encoding, :SUPER_SPECIAL_ENCODING]
irb(main):004:0> Encoding::Super_Special_Encoding
=> #<Encoding:Super Special Encoding>
irb(main):005:0> Encoding::Super_Special_Encoding.name
=> "Super Special Encoding"

And check this out:

irb(main):007:0> Encoding::UTF_8.replicate("\x7F")
=> #<Encoding:>
irb(main):008:0> Encoding::UTF_8.replicate("\x80")
(Object doesn't support #inspect)
=>                                                                                
irb(main):009:0> _.name
=> "\x80"
irb(main):010:0> _.encoding
=> #<Encoding:US-ASCII>
irb(main):011:0> Encoding.find("\x80")
(Object doesn't support #inspect)
=>                                                               
irb(main):012:0> _.name
=> "\x80"

Valid names apparently need to be asciicompat as per https://github.com/artichoke/artichoke/blob/trunk/artichoke-backend/vendor/ruby/include/ruby/internal/encoding/encoding.h#L766-L793 which is used here: https://github.com/artichoke/artichoke/blob/trunk/artichoke-backend/vendor/ruby/encoding.c#L288-L301

Eventually some more to look into maybe, but:

irb(main):001:0> constants = Encoding.constants
=> 
[:ISO8859_14,                                                          
...                                                                    
irb(main):002:0> (1..127).map(&:chr).each_slice(32).map { |c| Encoding::UTF_8.replicate(c.join) }
=> 
[#<Encoding:                                                           
                                                                       
                                                                       
 >,                                                                    
 #<Encoding:!"#$%&'()*+,-./0123456789:;<=>?@>,                         
 #<Encoding:ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`>,                         
 #<Encoding:abcdefghijklmnopqrstuvwxyz{|}~>]                           
irb(main):003:0> Encoding.constants - constants
=> [:ABCDEFGHIJKLMNOPQRSTUVWXYZ______, :ABCDEFGHIJKLMNOPQRSTUVWXYZ_____]

(Had to chunk it because otherwise some issue with Constants and lengths that I think it was getting mad at).

[b-n]

Note: in the between time, going to take the easy Bstr replace approach.

[lopopolo]

I think a better design is to pass encodings around as symbols (i.e. a u32) both in the interpreter storage, mrb_value repr for Encodings, and the flags of a String mrb_value.

If we end up doing that, we should use an FxHashSet here from rustc_hash which will perform much better