Figure better heuristic for proximity parameter in the JBR

[asn-d6]

We are currently using a heuristic based on the field size when computing the proximity parameter in the JBR.

There are likely better and cleaner heuristics or optimizations that we could use.

[hecmas]

I think we need to decide better how to choose the proximity parameter to both the JBR and UDR.

For example in the JBR @RogerTaule and myself observed the following.

With the current val: 1 - sqrt(rate) - gap we have that:

regime	total	ALI	DEEP	batching	commit round 1	commit round 2	commit round 3	commit round 4	commit round 5	query phase
UDR	53	186	167	163	169	173	177	181	184	53
JBR	49	185	166	49	75	99	123	142	160	63
best attack	128	—	—	—	—	—	—	—	—	—

This is the "best" case for the query phase, but notice that the security of the batching phase dropped to 49!

In the other hand if I set it to the "worst" case: (1 - rate)/2 + gap we have that:

regime	total	ALI	DEEP	batching	commit round 1	commit round 2	commit round 3	commit round 4	commit round 5	query phase
UDR	53	186	167	163	169	173	177	181	184	53
JBR	53	184	166	141	147	151	155	158	153	53
best attack	128	—	—	—	—	—	—	—	—	—

Which is "better" because now the minimum is 53 > 49. Moreover, contraty to batching, you can always improve it by doing more queries and/or adding grinding.

[asn-d6]

Pushed something here 8c32652
but it's very rough and rugged

/cc @asanso as well