Switch to new freebsd-unwind.h patch and add previously forgotten ldbl patch.

Sync ldbl patches with new revision sent to upstream for review.

Remove unrelated changes caused by makepatch.

https://cgit.freebsd.org/ports/commit/?id=4dfcc1b87d2f061d0daa1fc9fb7d81d16a86a255

Move include float.h to the top section.

It's ok, the only failure is arm64 GENERIC-UP kernel failure which is totally unrelated.

Remove LDBL_MANT_DIG guard from ld128 - those are only built on ld128
systems anyway.

Switch to using weak references.

Add all of the missing C23 _Float128 API functions.

1. Fix the date in UPDATING.
2. Correct arch(7). Settle on soft, because that's what the baseline will be.

@jhibbits
IMO this post https://www.talospace.com/2018/12/the-saga-of-power-isa-128-bit-long.html explains the issue properly.

In D57388#1320634, @jhibbits wrote:

In D57388#1320621, @pkubaj wrote:

I've just noticed that arch(7) should also be updated, but I'm not sure about one thing.

Architecture    float, double    long double
aarch64         hard             soft, quad precision
amd64           hard             hard, 80 bit
armv7           hard             hard, double precision
i386            hard             hard, 80 bit
powerpc         hard             hard, double precision
powerpcspe      hard             hard, double precision
powerpc64       hard             hard, double precision
powerpc64le     hard             hard, double precision
riscv64         hard             hard, quad precision

Obviously powerpc64le entry needs to be changed to quad. But I'm not really sure about whether it should be soft or hard. I mean, by default it will be soft, but if you build with CPUTYPE=power9 or newer, it's hard.

Are you saying that power8 doesn't have the instructions for hard quad, but power9 does? What instructions are required for this?

I've just noticed that arch(7) should also be updated, but I'm not sure about one thing.