[capture] Change PRNG for on-device test data generation (TVLA)

[johannheyszl]

We currently use Python random host-side and an implementation of the same Mersenne twister on device side.

• There is a danger of the Python class changing.
• This is not precisely in line with the definition of TVLA.
• Using something like AES to generate random data is very straightforward / comprehensible.

We could replace it by a SW implementation of AES to generate the data on-device.

[vrozic]

OT PR #20238 partially addressed this issue.
It adds AES SW implementation and changes the batch-capture commands to use AES generated data as defined in TVLA Derived Test Requirements.

After merging this, my plan to update the binaries and modify the capture scripts.

[m-temp]

The kmac script also depends on the random function.