Port bellard/quickjs "dtoa library" commits

[past-due]

Port bellard/quickjs dtoa library commits

added new dtoa library to print and parse float64 numbers. It is necessary to fix corner cases (e.g. radix != 10) and to have correct behavior regardless of the libc implementation.

Port of:
bellard/quickjs@9936606
bellard/quickjs@dbbca3d

Includes alternative fix for bellard/quickjs@638ec8c from 0191aea

Remaining TODO items (would appreciate a look @saghul @bnoordhuis):

• Amalgamation fails due to duplicate symbols (not sure how you'd prefer to fix this?)
• Codegen update This was actually a side effect of the root cause of the subsequent issue
• New test error 🤔 (see below - would appreciate a look at this one)

Result: 65/78606 errors, 3796 excluded, 5454 skipped, 2 new
test262/test/built-ins/TypedArrayConstructors/internals/Set/resized-out-of-bounds-to-in-bounds-index.js:18: unexpected error: Test262Error: Expected SameValue(«0», «100») to be true
test262/test/built-ins/TypedArrayConstructors/internals/Set/resized-out-of-bounds-to-in-bounds-index.js:18: strict mode: unexpected error: Test262Error: Expected SameValue(«0», «100») to be true

[bnoordhuis]

I have a fix for the first item in bnoordhuis/quickjs@2c073d7 that you can cherry-pick.

[saghul]

Can you run make gen? Looks like we're one bug away!

[past-due]

It turns out that I missed that the bellard/quickjs js_atof still used JS_NewFloat64 where quickjs-ng requires js_number() (since #319). Fixing that resolved the new test error (and also eliminated the codegen changes).

Something to keep a look out for in the future - any ports from bellard/quickjs that use JS_NewFloat64 should be double-checked.

[saghul]

Great work!

[past-due]

@bnoordhuis Anything else you're looking for to merge? (CI tests check out, and some local testing checks out fine as well.)

EDIT: Rebased on latest master