Enhance client to do reconnect on max connections from server

[tbeets]

The client will attempt reconnect when maximum connections exceeded error is sent by server immediately prior to a hard disconnect rather than close without reconnect attempt(s).

This is additive to existing re-connection on stale connection from server, auth and perm errors, and read loop errors (such as network disconnect).

[kozlovic]

For some reasons, the same tests are failing in push/PR.. locally they pass for me, wonder if this is just a coincidence or caused by this PR. Once you make the changes I requested, we will see if the same tests fail again.

[kozlovic]

@tbeets Discard my comment about same test failing in push/pr since you have only a PR run.. there were 2 tests (but different), one in 1.17 and one in 1.16 (the code coverage run). I am going to recycle to see if those were flappers or not.

[kozlovic]

hum... got this test failure again:

=== RUN   TestJetStreamClusterStreamLeaderChangeClientErr
    js_test.go:6334: Unexpected error: nats: raft: not leader
--- FAIL: TestJetStreamClusterStreamLeaderChangeClientErr (6.89s)

[coveralls]

Coverage increased (+0.04%) to 85.121% when pulling 06441fe on tbeets:main into e0e03e3 on nats-io:main.

[kozlovic]

LGTM

[wallyqs]

LGTM!