The original double boot protection code was written such that it assumed to be
executing synchronously before any events get processed. This assumption proved
to be incorrect since we were initializing the drivers before executing the ames
/boot scry.

It is somewhat unclear whether this mattered in practice. We discovered this by
investigating a ship that was peculiarly slow to to boot, but the root cause for
the slowness was the sponsoring galaxy having trouble.

We were also previously running the double boot protection check when doing the -X scry
which is certainly unnecessary.

See the parent commit for context. While the solution implemented in the parent
commit works, it is nasty and fragile because it relies on a halfway initialized
u3_pier. This refactors the solution to use the same mechanisms as the %wyrd
version negotiation, happening before _pier_work_init instead of in the middle
of it.

The original double boot protection code was written such that it
assumed to be executing synchronously before any events get processed.
This assumption proved to be incorrect since we were initializing the
drivers before executing the ames /boot scry.

It is somewhat unclear whether this mattered in practice. We discovered
this by investigating a ship that was peculiarly slow to to boot, but
the root cause for the slowness was the sponsoring galaxy having
trouble.

We were also previously running the double boot protection check when
doing the -X scry which is certainly unnecessary.

Otherwise zlib is fetched all the time for no reason, see ziglang/zig#23051 for details.

Otherwise zlib is fetched all the time for no reason, see
ziglang/zig#23051 for details.