doc/test-plan: create Blocksync testplan including various topologies including X bridge nodes and Y Full Nodes

[derrandz]

Context

We would like to benchmark how faster is this in comparison to previous syncing strategies (IPLD).

Facts to the moment:

• with IPLD was so slow that they never caught up (sync time > block time)
• IPLD will remain as a fallback in case of failure

For this benchmark, we will disable this fallback mechanism.

Topologies

The benchmark must simulate many sparsely connected FNs (“sparsely connected”: peerLimit = 3).
The benchmark must also simulate pretty big blocks.

Variations:

1. Keep only a few entry points from bootstrapping BNs to FNs

2. Variations on peer discovery for shrex:
   2.1 With discovery on
   - We can also very PeersLimit to see if it has an effect
   2.2 With discovery off
   2.3 With ONLY discovery (no full nodes)

3. Variations on type of syncing:
   3.1 Historical sync (blocks have already been produced)
   3.2 Syncing as blocks are being produced

4. Square sizes
   4.1 64
   4.2 128

Branch

At the moment, we will work off the following fork: (https://github.com/distractedm1nd/celestia-node/tree/shrex-integration-testing)

[distractedm1nd]

Correction to "With ONLY discovery": It isn't no full nodes, it just means there is no peers coming in through shrexsub

[distractedm1nd]

And lets pump up those square sizes!!! 256!!!