Treat the validator updates from the application as an unordered set

[ancazamfir]

When the application updates the validator set it calls updateValidators() with an 'updates' parameter, a list of changes that is then parsed in order, one entry at a time, making updates for the corresponding validator. i.e. the list is treated as a transaction list.
For example [add v1, rem v2, add v3] is different than [rem v2, add v1, add v3] and the call with one will produce different results than with the other.
This is an issue for gaia that requires these updates to be instead interpreted as a set.

[kevlubkcm]

sounds related to #3073

[cwgoes]

Proposal for atomic application:

• First apply all updates to the validator set, without setting priorities. Track a list of addresses of newly added validators.
• Calculate the total voting power of the new validator set.
• Set the priorities of all newly added validators to -C * total voting power (presently C = 1.125).
• Normalize priorities if necessary & subtract the average.

Also we should consider if we need C = 1.125 after this change.

[ancazamfir]

Proposal for atomic application:
First apply all updates to the validator set, without setting priorities. Track a list of addresses of newly added validators.

all updates except the validator removals

Calculate the total voting power of the new validator set.
Set the priorities of all newly added validators to -C * total voting power (presently C = 1.125).

do validator removals here

Normalize priorities if necessary & subtract the average.
Also we should consider if we need C = 1.125 after this change.

[ancazamfir]

pls. see pull request #3222

[ancazamfir]

Seems like NewValidatorSet(valList) does not check for duplicates in the validator list valList and we end up with a validator "set" including same validator multiple times. I'm not sure if all checks are in place when running in a real setup so valList does not contain duplicates. Will check.
This also needs fix it in validator_set.go

[ebuchman]

Merged! Thanks @ancazamfir ! Will add follow up notes to #3166