Input format on witness generation

[damesca]

I have found a detail concerning the input data for witness generation.

There are two ways to enter numbers:
(1) as json numbers, e.g., "a": 1234,
(2) as json strings, e.g., "a": "1234".

Both ways work fine with small numbers, but when the input is a bigint, it must be passed in string format, otherwise precision is lost and the input is internally modified, leading to incorrect results.

It would be a useful detail for newcomers to have this specified in the README, in section 14 ("calculate the witness").