arm/linux: zero-initialize stack buffers to fix Valgrind warnings

[vozvivan]

Description

This PR fixes Valgrind and MemorySanitizer warnings caused by reading uninitialized stack memory.

1. Bugfix:
On modern AArch64 kernels, the "Hardware" line is often completely absent from /proc/cpuinfo. When this happens, the stack-allocated proc_cpuinfo_hardware buffer is never populated and may remains uninitialized garbage. Passing it to strnlen and strncmp in cpuinfo_arm_linux_decode_chipset_from_proc_cpuinfo_hardware triggers memory analyzer warnings. Adding = {0} ensures it safely defaults to an empty string "" and fixes the issue.

2. Follow-up hardening:
To proactively prevent the exact same issue on other platforms (or Android) where OS-specific properties might be missing, I also zero-initialized the adjacent proc_cpuinfo_revision and android_properties buffers.

Since cpuinfo_arm_linux_init runs exactly once via pthread_once, the performance overhead of zeroing these bytes is negligible.

Valgrind Trace

==24496== Conditional jump or move depends on uninitialised value(s)
==24496==    at 0x4838B0C: strnlen (vg_replace_strmem.c:468)
==24496==    by 0x1B3CCFF: cpuinfo_arm_linux_decode_chipset_from_proc_cpuinfo_hardware
==24496==    by 0x1B3E24B: cpuinfo_arm_linux_decode_chipset
==24496==    by 0x1B3AC47: cpuinfo_arm_linux_init

[GregoryComer]

Closing and re-opening to fix stuck CI.

This PR was reopened (likely due to being reverted), so your approval was removed. Please request another review.