Branch to global symbol results in reference to PLT, and when compiling
for THUMB-2 - in a R_ARM_THM_JUMP19 relocation. Some linkers don't
support this relocation (ld.gold), while others can end up truncating
the relocation to fit (ld.bfd).

Convert this branch through PLT into a direct branch that the assembler
can resolve locally.

See https://github.com/android-ndk/ndk/issues/337 for background.

The current workaround is to disable poly1305 optimization assembly,
which is not optimal and can be reverted after this patch:
https://github.com/freedesktop/gstreamer-cerbero/commit/beab607d2b1ff23c41b7e01aa9c64be5e247d1e6

CLA: trivial
Reviewed-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5949)

Reviewed-by: NRichard Levitte <levitte@openssl.org>

Even though no test could be found to trigger this, paper-n-pencil
estimate suggests that x86 and ARM inner loop lazy reductions can
loose a bit in H4>>*5+H0 step.
Reviewed-by: NEmilia Käsper <emilia@openssl.org>

RT#4483

[poly1305-armv4.pl: remove redundant #ifdef __thumb2__]
[poly1305-ppc*.pl: presumably more accurate benchmark results]
Reviewed-by: NRichard Levitte <levitte@openssl.org>

Some of these scripts would recognise an output parameter if it looks
like a file path.  That works both in both the classic and new build
schemes.  Some fo these scripts would only recognise it if it's a
basename (i.e. no directory component).  Those need to be corrected,
as the output parameter in the new build scheme is more likely to
contain a directory component than not.
Reviewed-by: NAndy Polyakov <appro@openssl.org>

Formally only 32-bit AVX2 code path needs this, but I choose to
harmonize all vector code paths.

RT#4346
Reviewed-by: NRichard Levitte <levitte@openssl.org>

Most of the assembly uses constants from arm_arch.h, but a few references to
ARMV7_NEON don't. Consistently use the macros everywhere.
Signed-off-by: NAndy Polyakov <appro@openssl.org>
Reviewed-by: NRich Salz <rsalz@openssl.org>

Different assembler versions disagree on how to interpret #-1 as
argument to vmov.i64, as 0xffffffffffffffff or 0x00000000ffffffff.
So replace it with something they can't disagree on.
Reviewed-by: NRich Salz <rsalz@openssl.org>

Reviewed-by: NRichard Levitte <levitte@openssl.org>