More unique uniqid (no BC)

[marc-mabe]

Currently uniqid waits until the current microsecond has been changed until the previous call.

This PR changes this logic instead of waiting until the current microsecond changes, it simply increments in case the current time is <= the previous.

This makes uniqid unaffected by time changes (like ntp) and also it's faster (if called very often) due to less system calls.

A simple benchmark:
BEFORE: 100911477ns
AFTER: 57341969ns

uniqid(); // warmup

$start = hrtime(true);
for ($i = 0; $i < 100000; ++$i) {
        uniqid();
}
echo (hrtime(true) - $start) . "ns\n";

[TimWolla]

This effectively is a duplicate of #18232.

[marc-mabe]

This effectively is a duplicate of #18232.

OMG, I wasn't aware there is already another PR addressing the same.
Also, if this requires an RFC I'll not continue on that as this function should be avoided anyway.