improve grouping with self-loops

[adisos]

          ```suggestion

vsi3a-ky[10.240.30.5],vsi3b-ky[10.240.30.6],vsi3c-ky[10.240.30.4] => vsi3a-ky[10.240.30.5],vsi3b-ky[10.240.30.6],vsi3c-ky[10.240.30.4] : All Connections
vsi3a-ky[10.240.30.5],vsi3b-ky[10.240.30.6],vsi3c-ky[10.240.30.4] => vsi1-ky[10.240.10.4] : All Connections

This requires adding a self loop to each VSI, run the grouping algorithm, then remove left-over self loops.
Can be handled in a future PR.

_Originally posted by @zivnevo in https://github.com/np-guard/vpc-network-config-analyzer/pull/92#discussion_r1253989708_
            

[ShiriMoran]

This requires adding a self loop to each VSI, run the grouping algorithm, then remove left-over self loops.
Can be handled in a future PR.

_Originally posted by @zivnevo in https://github.com/np-guard/vpc-network-config-analyzer/pull/92#discussion_r1253989708_
            

@zivnevo adding a self loop before running the algorithm seems wrong.
Consider subnets a,b,c,d with the following connections:

a => b
a => c
d => c
d => b

With the current algorithm, after grouping src we'll get

a => b,c
d => b,c

and after grouping dst on the result we'll get
a,d => b,c

but if we add self loops before running the alg after grouping src we'll get:

a => a,b,c
d => b,c,d

and grouping dst will group nothing.

I'm working on a different solution.

[zivnevo]

My initial thought was to try both with and without self-loops and see which gives better results. Another option is to add the self loops just after some initial grouping was done.
If you have a one-shot idea, this might be better. Please share you algorithm here.

[ShiriMoran]

Actually I'm afraid that if we choose the direction of adding self loops, grouping and comparing than what is really required is to check all the possible combinations of subnets/vsis with and without self loops, which is not reasonable for a couple of reasons.
E.g., consider the following example:

a => b
a => c
b => c

To gain the optimal grouping a self loop should be added only to b, which yields
a, b => b, c

I have another solution in mind, if you wish we can discuss or I can put it here before I write the code.

[ShiriMoran]

Proposed algorithm:

• execute the current grouping algorithm for grouping, assume w.l.o.g we grouped destinations
• For each two lines, compute their distance
• merge all lines whose distance is an empty set; the claim below guarantees that the result is coherent

The distance between two GroupedConnLine is defined as following:
Let l_1 be a line with source s_1 and dest d_1 and let l_2 be a line with source s_2 and dest d_2.
l_1 / l_2 is the subnets in d_1 that are not in d_2 minus the single subnet in s_1 if |s_1| = 1

The distance between lines l_1 and l_2 is l_1 / l_2 union l_2 / l_2

claim: if the distance between line l_1 and l_2 is empty and the distance between lines l_2 and l_3 is zero then so is the distance between l_1 and l_3

[ShiriMoran]

solved by #152