Lexer deficiencies with numbers

[sauclovian-g]

This does not strike me as user-friendly behavior:

sawscript> :t 0b1
Int
sawscript> :t 0b2

Type errors:
  <stdin>:1:1-1:2: Type mismatch.
    Term is not a function. (Maybe a function is applied to too many arguments?)
    <stdin>:1:1-1:2: The type t.0 -> t.1 arises from the context of the term
    <stdin>:1:1-1:2: The type Int arises from the type of this term
    
    Expected: t.0 -> t.1
    Found:    Int
    
    within "_" (<stdin>:1:1-1:4)

  <stdin>:1:2-1:4: Unbound variable: "b2" (<stdin>:1:2-1:4)
Note that some built-in commands are available only after running
either `enable_deprecated` or `enable_experimental`.

Observe that it treats 0b2 as 0 followed by the identifier b2 rather than as a malformed binary constant. Similar things happen with 0xg and 0o8.

I think it would be better to match [digit]alnum* as number first and then reject malformed ones, as that's what people generally expect in this day and age.

[sauclovian-g]

On a similar front, string constants with malformed contents are dropped by the lexer, resulting in a confusing error message:

sawscript> return "abc\tdef"
[03:05:18.731] "abc\tdef"
sawscript> return "abc\qdef"
Parse error at <stdin>:1:8-1:9: Unexpected `"'