Try to free some disk space on the arm64 runners (osquery#7950)

Set counter consistently so zero always indicates all records (osquer…

…y#7801)

cve: Ignore zstd CVE-2021-24031 (osquery#7865)

Improvements to osquery AWS logic (osquery#7714)

- Introduce the flag aws_imdsv2_request_attempts
  to specify how many attempts should be done
  to retrieve an IMDSv2 token to do a secure request.

- Introduce the flag aws_imdsv2_request_interval
  to specify the base seconds to wait between attempts,
  which scales quadratically with the number of attempts.

- Introduce the flag aws_disable_imdsv1_fallback
  which disables IMDSv1 as a fallback if the IMDSv2 token
  fails to be retrieved.

- Remove the automatic check to see if osquery
  is running on an EC2 instance.

- Improve the retrieval of instance id and region.
  If the retrieval keeps failing, don't cache empty values,
  keep retrying on next requests until it has success,
  then cache the values.

- Improve error message when STS credentials fail to be retrieved.
  The hardcoded error was hiding the true reason for the failure.

libs: Update sqlite to version 3.39.2 (osquery#7736)

libs: Update libxml2 to v2.9.14 (osquery#7729)

expand env vars for the windows_crashes table (osquery#7391)

Update `osquery.example.conf` to omit the CLI only flags (osquery#7595)

Fix user_time and system_time unit in processes table on M1 (osquery#…

…7473)

Use the time base conversion coming from mach_timebase_info
also for the cpu time values coming from the libproc API.

The user_time and system_time columns values are taken
from the proc_pid_rusage API values of ri_user_time and
ri_system_time.
The unit of those two values was incorrectly considered
to always be nanoseconds, but they are mach ticks,
much like the start_time that is later calculated
via mach_absolute_time.

While on x86_64 the time base info used to convert ticks to nanoseconds
returns a numerator and denominator that are both 1,
on M1 they are not (125 and 3 is one example).