Compilation: strip debug info from ReleaseSmall by default

[alichraghi]

Closes #6676

ali@linux ~/d/zig (o0)> build/zig2 build-exe z.zig -O ReleaseSmall
ali@linux ~/d/zig (o0)> ls -lha ./z ./z.o
-rwxr-xr-x 1 ali ali 8.5K Oct  4 17:52 ./z*
-rw-r--r-- 1 ali ali  12K Oct  4 17:52 ./z.o
ali@linux ~/d/zig (o0)> build/zig2 build-exe z.zig -O ReleaseSmall --strip
ali@linux ~/d/zig (o0)> ls -lha ./z ./z.o
-rwxr-xr-x 1 ali ali 8.5K Oct  4 17:52 ./z*
-rw-r--r-- 1 ali ali  12K Oct  4 17:52 ./z.o
ali@linux ~/d/zig (o0)> build/zig2 build-exe z.zig -O ReleaseSmall --no-strip
ali@linux ~/d/zig (o0)> ls -lha ./z ./z.o
-rwxr-xr-x 1 ali ali 426K Oct  4 17:53 ./z*
-rw-r--r-- 1 ali ali 729K Oct  4 17:53 ./z.o
ali@linux ~/d/zig (o0)> build/zig2 build-exe z.zig
ali@linux ~/d/zig (o0)> ls -lha ./z ./z.o
-rwxr-xr-x 1 ali ali 976K Oct  4 17:53 ./z*
-rw-r--r-- 1 ali ali 1.4M Oct  4 17:53 ./z.o

[wooster0]

Also, Andrew suggested -fstrip and -fno-strip instead of --strip and --no-strip (#6676 (comment)); have you considered that?

[alichraghi]

thanks. i missed that

[alichraghi]

Now, will it make sense to disable strip by default for zig test or zig build test?
* by "default", i mean it's null so it get enabled in ReleaseSmall

[Vexu]

Whoops, looks like the Windows failure wasn't spurious but caused by a need for a new tarball (sigh). #13055 also needs one so maybe these two should be handled together.

[Luukdegram]

Looks like this also caused the CI to error out for the linker test for the 'producers' section.
Fixed this in 8635c18