release-23.1.20-rc: stats: more conservative lower bound for downward-trending forecasts

[michae2]

Backport:

• 1/1 commits from "stats: more conservative lower bound for downward-trending forecasts" (stats: more conservative lower bound for downward-trending forecasts #122273)
• 1/1 commits from "sql/stats: turn forecasting constants into cluster settings" (sql/stats: turn forecasting constants into cluster settings #122779)

/cc @cockroachdb/release

---

stats: more conservative lower bound for downward-trending forecasts

Instead of 0 as a lower bound for predictions, use 1/3 times the lowest prior observation. This will avoid predicting zero rows for a statistic unless we really observed zero rows somewhat recently.

Fixes: #119967

Release note (bug fix): Statistics forecasts of zero rows can cause bad plans. This commit changes forecasting to avoid predicting zero rows for most downward-trending statistics.

sql/stats: turn forecasting constants into cluster settings

Turn three constants used for statistics forecasting into cluster settings. (These must be cluster settings rather than session variables because forecasting happens when reloading table statistics in the stats cache, which can be independent of any session.)

Informs: #119967

Release note (sql change): Introduce three new cluster settings for controlling table statistics forecasting:

1. sql.stats.forecasts.min_observations is the minimum number of observed statistics required to produce a forecast.

2. sql.stats.forecasts.min_goodness_of_fit is the minimum R² (goodness of fit) measurement required from all predictive models to use a forecast.

3. sql.stats.forecasts.max_decrease is the most a prediction can decrease, expressed as the minimum ratio of the prediction to the lowest prior observation.

---

Release justification: specially-requested fix for plan regression bug.

[blathers-crl]

Thanks for opening a backport.

Please check the backport criteria before merging:

• Backports should only be created for serious
  issues or test-only changes.
• Backports should not break backwards-compatibility.
• Backports should change as little code as possible.
• Backports should not change on-disk formats or node communication protocols.
• Backports should not add new functionality (except as defined
  here).
• Backports must not add, edit, or otherwise modify cluster versions; or add version gates.
• All backports must be reviewed by the owning areas TL and one additional
  TL. For more information as to how that review should be conducted, please consult the backport
  policy.

If your backport adds new functionality, please ensure that the following additional criteria are satisfied:

• There is a high priority need for the functionality that cannot wait until the next release and is difficult to address in another way.
• The new functionality is additive-only and only runs for clusters which have specifically “opted in” to it (e.g. by a cluster setting).
• New code is protected by a conditional check that is trivial to verify and ensures that it only runs for opt-in clusters. State changes must be further protected such that nodes running old binaries will not be negatively impacted by the new state (with a mixed version test added).
• The PM and TL on the team that owns the changed code have signed off that the change obeys the above rules.
• Your backport must be accompanied by a post to the appropriate Slack
  channel (#db-backports-point-releases or #db-backports-XX-X-release) for awareness and discussion.

Also, please add a brief release justification to the body of your PR to justify this
backport.

[cockroach-teamcity]

This change is 

[yuzefovich]

Reviewed 3 of 3 files at r1, 4 of 4 files at r2, all commit messages.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @mgartner)

[michae2]

Fixes for the tests coming up, but this is RFAL.

[mgartner]

Reviewed 3 of 3 files at r1, 4 of 4 files at r3, 5 of 5 files at r4, all commit messages.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @michae2)

[mgartner]

Should we set this to 1 by default in backports to retain the old behavior? Would that reduce risk of this change?

[michae2]

Good point, I will do that.

[rytaft]

Reviewed 3 of 3 files at r1, 4 of 4 files at r3, 5 of 5 files at r5, all commit messages.
Reviewable status: complete! 1 of 0 LGTMs obtained (and 1 stale) (waiting on @mgartner)