executor: fix panic when fetching processlist table data in some cases

[qw4990]

What problem does this PR solve?

When TiDB fetches data from processlist table, it will get each connection's memory consumption according to their Ctx.Memtracker.
But in some cases, Ctx.Memtracker is not initialized at that time, which results in a panic.

What is changed and how it works?

Check if Ctx.Memtracker is nil before using it when fetching processlist table data.

[codecov]

Codecov Report

Merging #11688 into master will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff             @@
##             master     #11688   +/-   ##
===========================================
  Coverage   81.4407%   81.4407%           
===========================================
  Files           429        429           
  Lines         92746      92746           
===========================================
  Hits          75533      75533           
  Misses        11821      11821           
  Partials       5392       5392

[lysu]

LGTM

[zz-jason]

Could you add some UT?

[jackysp]

LGTM

[qw4990]

Could you add some UT?

It's hard to write some unit tests for this case.
The state of a Conn is complex so it's hard to construct an appropriate Conn to trigger this bug.

Actually, to locate this problem, I create an individual goroutine for each Conn when it comes.
This individual goroutine executes a dead loop to find the invalid state of this Coon:

for {
    If conn.ProcessInfo() != nil && conn.Ctx.MemTracker == nil {
        panic("invalid state")
    }
}

It's not appropriate to write dead loops in our unit tests.
PTAL. @zz-jason

[qw4990]

PTAL @zz-jason

[zz-jason]

LGTM

[sre-bot]

/run-all-tests

[sre-bot]

cherry pick to release-3.0 failed

[sre-bot]

It seems that, not for sure, we failed to cherry-pick this commit to release-3.0. Please comment '/run-cherry-picker' to try to trigger the cherry-picker if we did fail to cherry-pick this commit before. @qw4990 PTAL.