Bit lengths seem to be computed incorrectly

[cjpatton]

We compute the bit length for the squared norm and wraparound range checks as math.ceil(math.log2(m)) where m is the largest, valid value:

https://github.com/junyechen1996/draft-chen-cfrg-vdaf-pine/blob/21c43447f9b3ed283cc44500001ab4e9411a72c7/poc/flp_pine.py#L82

https://github.com/junyechen1996/draft-chen-cfrg-vdaf-pine/blob/21c43447f9b3ed283cc44500001ab4e9411a72c7/poc/flp_pine.py#L101

This appears to be incorrect when m is a power of 2:

m	bit encoding	bit length
1	1	1
2	10	2
4	100	3
8	1000	4
16	10000	5

What we want instead is math.floor(math.log2(m)+1).

Thanks to @jhoyla for pointing out this bug.

[junyechen1996]

Both look correct to me?

m = 1, math.ceil(math.log2(m + 1)) = 1, math.floor(math.log2(m) + 1) = 1 m = 2, math.ceil(math.log2(m + 1)) = 2, math.floor(math.log2(m) + 1) = 2 m = 4, math.ceil(math.log2(m + 1)) = 3, math.floor(math.log2(m) + 1) = 3

[cjpatton]

Ohhh right, the +1 is important.