Reduce the committee size as a way to reduce the bar for power participation percent

[masih]

Came up in 1:1 with Molly:

• We already use the scaled power to determine participation in f3 (65535)
• What would happen if the power is scaled over a smaller value such that the effective committee size is reduced per instance.
• This does not directly reduce the need for higher participation power percent, but reduces the number of participants.
  • The participants left would have higher power portion over the whole network and hopefully better connectivity.
• The catch of course is higher sensitivity to nodes with more power.

[masih]

What we probably should is to make this a manifest param for experimentation

[Stebalien]

This gets a bit tricky because SPs have multiple miner actors. If we did this, we'd likely want delegation of power which would, IMO, fix the issue entirely. Unfortunately, delegation leads to centralization which may not be a huge issue given that we're only finalizing blocks but I know @Kubuxu is not a fan of this.