Remove uneeded PTX code generation from libfaiss builds.

The CMake CUDA Architecture value of `60` means to generate
both PTX and SASS for that arch. We only need SASS for the
architectures we support, and one PTX version for future
hardware versions.

So now we build on SASS for everything ( `60-real` ) and
use 80 as the baseline for newer archs likes 90

By removing this unneeded PTX code we can reduce
the libfaiss.a binary to 305MB from the current 484MB.

.circleci/config.yml

@@ -350,7 +350,7 @@ workflows:
           exec: linux-x86_64-gpu
           label: main
           cuda: "11.4"
-          cuda_archs: "60;61;70;72;75;80;86"
+          cuda_archs: "60-real;61-real;62-real;70-real;72-real;75-real;80;86-real"
           compiler_version: "11.2"
           filters:
             tags:
@@ -363,7 +363,7 @@ workflows:
           label: main
           raft: "ON"
           cuda: "11.4"
-          cuda_archs: "60;61;70;72;75;80;86"
+          cuda_archs: "60-real;61-real;62-real;70-real;72-real;75-real;80;86-real"
           compiler_version: "11.2"
           filters:
             tags:
@@ -415,15 +415,15 @@ workflows:
           name: Linux x86_64 GPU nightlies (CUDA 11.4)
           exec: linux-x86_64-gpu
           cuda: "11.4"
-          cuda_archs: "60;61;70;72;75;80;86"
+          cuda_archs: "60-real;61-real;62-real;70-real;72-real;75-real;80;86-real"
           compiler_version: "11.2"
           label: nightly
       - build_conda:
           name: Linux x86_64 GPU w/ RAFT nightlies (CUDA 11.4)
           exec: linux-x86_64-gpu
           raft: "ON"
           cuda: "11.4"
-          cuda_archs: "60;61;70;72;75;80;86"
+          cuda_archs: "60-real;61-real;62-real;70-real;72-real;75-real;80;86-real"
           compiler_version: "11.2"
           label: nightly
       - build_conda: