optimizing is_ipv4

[lemire]

This PR optimizes the is_ipv4 function by first checking the last character. Unless the last character is a hex number or the letter x, we can just return false immediately and save quite a bit of work.

The function was also rewritten with more safeguards.

LLVM 14, Apple M2.

Before:

BasicBench_AdaURL_href              24053490 ns     24053414 ns           29 GHz=3.38364 cycle/byte=9.37729 cycles/url=814.504 instructions/byte=38.9054 instructions/cycle=4.14889 instructions/ns=14.0383 instructions/url=3.37929k ns/url=240.719 speed=361.2M/s time/byte=2.76855ns time/url=240.474ns url/s=4.15845M/s
BasicBench_AdaURL_aggregator_href   16111354 ns     16111372 ns           43 GHz=3.48936 cycle/byte=6.19904 cycles/url=538.444 instructions/byte=27.773 instructions/cycle=4.48021 instructions/ns=15.6331 instructions/url=2.41234k ns/url=154.31 speed=539.252M/s time/byte=1.85442ns time/url=161.073ns url/s=6.20835M/s

After

BasicBench_AdaURL_href              24135053 ns     24135033 ns           30 GHz=3.39617 cycle/byte=9.33507 cycles/url=810.837 instructions/byte=38.7939 instructions/cycle=4.15571 instructions/ns=14.1135 instructions/url=3.3696k ns/url=238.751 speed=359.978M/s time/byte=2.77794ns time/url=241.29ns url/s=4.14439M/s
BasicBench_AdaURL_aggregator_href   15683330 ns     15683273 ns           44 GHz=3.3813 cycle/byte=6.09465 cycles/url=529.377 instructions/byte=27.3717 instructions/cycle=4.49111 instructions/ns=15.1858 instructions/url=2.37749k ns/url=156.56 speed=553.972M/s time/byte=1.80515ns time/url=156.794ns url/s=6.37781M/s

[anonrig]

I think we have a bug in here. This is somewhat similar to nodejs/node#50632

D:\a\ada\ada\tests\wpt_tests.cpp(385): error: Expected equality of these values:
  input_url->get_host()
    Which is: "256"
  host
    Which is: "0.0.1.0"

[lemire]

@anonrig Seems unrelated to this Node issue unless I misunderstand something. I have a Windows box open in front of me and I am failing to reproduce this issue.

[lemire]

The failing test is this...

{
    "input": "http://256",
    "base": "http://other.com/",
    "href": "http://0.0.1.0/",
    "origin": "http://0.0.1.0",
    "protocol": "http:",
    "username": "",
    "password": "",
    "host": "0.0.1.0",
    "hostname": "0.0.1.0",
    "port": "",
    "pathname": "/",
    "search": "",
    "hash": ""
  }

But this test is not new. And I cannot reproduce it locally.

And it is only triggered by Visual Studio in release mode. And only in the last few commits.

Weird.

What could it be?

[lemire]

@anonrig I am merging this. We can revert later if you disagree.