Fix panic of adjacent-region-scheduler after pd transfer leader

[nanne007]

What problem does this PR solve?

If adjacent-region scheduler is enabled, after pd transfers leader and the new leader is prepared,
there is a gap that heartbeats of some regions are not received, and leaders are not updated.

store id of a nil peer is 0, see https://github.com/pingcap/pd/blob/fa5f08c252a4c5025cdef5bf0f2bdab78d7c0a36/vendor/github.com/pingcap/kvproto/pkg/metapb/metapb.pb.go#L256, so this line https://github.com/pingcap/pd/blob/fa5f08c252a4c5025cdef5bf0f2bdab78d7c0a36/server/schedulers/adjacent_region.go#L223 return a nil store, which causes panic.

What is changed and how it works?

This change makes sure that region's leader is not nil, and leader's store is not nil, and also fix metric value of adjacent-count.

Another solution is skip holes of not heartbeat-ed regions when calculating adjacent-regions, which prevents the issue at front.

Check List

Tests

make test

No side effects

[sre-bot]

Hi contributor, thanks for your PR.

This patch needs to be approved by someone of admins. They should reply with "/ok-to-test" to accept this PR for running test automatically.

[disksing]

/ok-to-test

[disksing]

PTAL @nolouch

[disksing]

LGTM.

[nolouch]

LGTM.