crypto/rsa: allocate nats on the stack for RSA 2048

crypto/rsa: allocate nats on the stack for RSA 2048

With a small tweak and the help of the inliner, we preallocate enough
nat backing space to do RSA-2048 on the stack.

We keep the length of the preallocated slices at zero so they don't
silently mask missing expandFor calls.

Surprisingly enough, this doesn't move the CPU benchmark needle much,
but probably reduces GC pressure on larger applications.

name                    old time/op    new time/op    delta
DecryptPKCS1v15/2048-8    1.25ms ± 0%    1.22ms ± 1%   -1.68%  (p=0.000 n=10+9)
DecryptPKCS1v15/3072-8    3.78ms ± 0%    3.73ms ± 1%   -1.33%  (p=0.000 n=9+10)
DecryptPKCS1v15/4096-8    8.62ms ± 0%    8.45ms ± 1%   -1.98%  (p=0.000 n=8+10)
EncryptPKCS1v15/2048-8     140µs ± 1%     136µs ± 0%   -2.43%  (p=0.000 n=9+9)
DecryptOAEP/2048-8        1.25ms ± 0%    1.24ms ± 0%   -0.83%  (p=0.000 n=8+10)
EncryptOAEP/2048-8         140µs ± 0%     137µs ± 0%   -1.82%  (p=0.000 n=8+10)
SignPKCS1v15/2048-8       1.29ms ± 0%    1.29ms ± 1%     ~     (p=0.574 n=8+8)
VerifyPKCS1v15/2048-8      139µs ± 0%     136µs ± 0%   -2.12%  (p=0.000 n=9+10)
SignPSS/2048-8            1.30ms ± 0%    1.28ms ± 0%   -0.96%  (p=0.000 n=8+10)
VerifyPSS/2048-8           140µs ± 0%     137µs ± 0%   -1.99%  (p=0.000 n=10+8)

name                    old alloc/op   new alloc/op   delta
DecryptPKCS1v15/2048-8    15.0kB ± 0%     0.5kB ± 0%  -96.58%  (p=0.000 n=10+10)
DecryptPKCS1v15/3072-8    24.6kB ± 0%     3.3kB ± 0%  -86.74%  (p=0.000 n=10+10)
DecryptPKCS1v15/4096-8    38.9kB ± 0%     4.5kB ± 0%  -88.50%  (p=0.000 n=10+10)
EncryptPKCS1v15/2048-8    18.0kB ± 0%     1.2kB ± 0%  -93.48%  (p=0.000 n=10+10)
DecryptOAEP/2048-8        15.2kB ± 0%     0.7kB ± 0%  -95.10%  (p=0.000 n=10+10)
EncryptOAEP/2048-8        18.2kB ± 0%     1.4kB ± 0%  -92.29%  (p=0.000 n=10+10)
SignPKCS1v15/2048-8       21.9kB ± 0%     0.8kB ± 0%  -96.50%  (p=0.000 n=10+10)
VerifyPKCS1v15/2048-8     17.7kB ± 0%     0.9kB ± 0%  -94.85%  (p=0.000 n=10+10)
SignPSS/2048-8            22.3kB ± 0%     1.2kB ± 0%  -94.77%  (p=0.000 n=10+10)
VerifyPSS/2048-8          17.9kB ± 0%     1.1kB ± 0%  -93.75%  (p=0.000 n=10+10)

name                    old allocs/op  new allocs/op  delta
DecryptPKCS1v15/2048-8       124 ± 0%         3 ± 0%  -97.58%  (p=0.000 n=10+10)
DecryptPKCS1v15/3072-8       140 ± 0%         9 ± 0%  -93.57%  (p=0.000 n=10+10)
DecryptPKCS1v15/4096-8       158 ± 0%         9 ± 0%  -94.30%  (p=0.000 n=10+10)
EncryptPKCS1v15/2048-8      80.0 ± 0%       7.0 ± 0%  -91.25%  (p=0.000 n=10+10)
DecryptOAEP/2048-8           130 ± 0%         9 ± 0%  -93.08%  (p=0.000 n=10+10)
EncryptOAEP/2048-8          86.0 ± 0%      13.0 ± 0%  -84.88%  (p=0.000 n=10+10)
SignPKCS1v15/2048-8          162 ± 0%         4 ± 0%  -97.53%  (p=0.000 n=10+10)
VerifyPKCS1v15/2048-8       79.0 ± 0%       6.0 ± 0%  -92.41%  (p=0.000 n=10+10)
SignPSS/2048-8               167 ± 0%         9 ± 0%  -94.61%  (p=0.000 n=10+10)
VerifyPSS/2048-8            84.0 ± 0%      11.0 ± 0%  -86.90%  (p=0.000 n=10+10)

Change-Id: I511a2f5f6f596bbec68a0a411e83a9d04080d72a
Reviewed-on: https://go-review.googlesource.com/c/go/+/445021
Run-TryBot: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Joedian Reid <joedian@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>