Calculate vnodes on daprd instead of the placement service

[elena-kolevska]

Description

(Note: Please refer to the issue for the reason behind this PR)

This PR removes the vnodes and their sorted set from the placement table that's disseminated from the placement service to the dapr sidecars, and instead, it calculates them on the dapr sidecars.
It accounts for mismatch of versions in daprd/placement services (older daprd version with placement on 1.13+ and vice versa). In the next Dapr version we should be able to remove these checks and clean up the code.

A few notes that might help during the review:

• replicationFactor is an argument received on placement service startup, whose default value is 100.
• We pass the replication factor as part of the placement table message to daprd, so we can calculate the vnodes there
• We attach the api level of the daprd as metadata to the stream connection. The placement service will look at this data and decide if it needs to send the vnodes based on it (needed for daprd sidecars on older versions). This way we're not relying on the minimum API level of the cluster, but serve data based on what the particular sidecar needs.
• We still need to keep the vnodes in state on the placement side because of backwards compatibility.

Issue reference

#7360

Checklist

Please make sure you've completed the relevant tasks for this PR, out of the following list:

• Code compiles correctly
• Created/updated tests
• Unit tests passing
• End-to-end tests passing

N/A

• Extended the documentation / Created issue in the https://github.com/dapr/docs/ repo: dapr/docs#[issue number]
• Specification has been updated / Created issue in the https://github.com/dapr/docs/ repo: dapr/docs#[issue number]
• Provided sample for the feature / Created issue in the https://github.com/dapr/docs/ repo: dapr/docs#[issue number]

[codecov]

Codecov Report

Attention: 41 lines in your changes are missing coverage. Please review.

Comparison is base (f9a2088) 62.36% compared to head (db1f17c) 62.34%.
Report is 1 commits behind head on master.

Files	Patch %	Lines
pkg/placement/hashing/consistent_hash.go	48.83%	21 Missing and 1 partial ⚠️
pkg/placement/membership.go	46.42%	9 Missing and 6 partials ⚠️
pkg/placement/placement.go	62.50%	1 Missing and 2 partials ⚠️
pkg/actors/placement/placement.go	92.85%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #7382      +/-   ##
==========================================
- Coverage   62.36%   62.34%   -0.03%     
==========================================
  Files         240      240              
  Lines       22075    22154      +79     
==========================================
+ Hits        13767    13811      +44     
- Misses       7156     7186      +30     
- Partials     1152     1157       +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[artursouza]

This PR warms my heart because it addresses an obvious problem that nobody dared to point out AND fix. Thanks a lot.

[artursouza]

Why the old name is used for the new behavior? Am I missing something?

[elena-kolevska]

So my thinking was that we rename the old behaviour to NewFromExistingWithVirtNodes and then we remove it in the next version. And we keep the new behaviour with a name that's gonna read more naturally in this version, but also in the next ones.

[elena-kolevska]

I added caching for the virtual hosts. I'll be adding a few more tests today and also I'll fix the Test_Integration/placement/apilevel/noMax/run test that I seem to have broken.

Addressed