* test/e2e/operator_test.go (waitForDeploymentEnvVar): Use the provided
timeout parameter instead of a hard-coded value of 1 minute.

Specify a time unit suffix "s" for seconds in the value of the
RELOAD_INTERVAL environment variable in router deployments.  Omitting the
time unit causes the following warning:

    router "msg"="invalid interval, using default"  "default"=5 "interval"="5" "name"="RELOAD_INTERVAL"

* pkg/operator/controller/ingress/deployment.go (desiredRouterDeployment):
Add "s" time unit suffix to the reload interval.
* pkg/operator/controller/ingress/deployment_test.go
(TestDesiredRouterDeployment):
* test/e2e/operator_test.go (TestReloadIntervalUnsupportedConfigOverride):
Expect the time unit suffix.

Add an unsupported config override for setting HAProxy's global maxconn
setting.  If the config override is unspecified, the current default of
20000 is used.  If the config override has the value "-1", the maxconn
setting is omitted so that HAProxy automatically detects a reasonable value
based on the value of ulimit -n.

* pkg/operator/controller/ingress/deployment.go
(RouterMaxConnectionsEnvName): New const.
(desiredRouterDeployment): Add unsupported config override for
ROUTER_MAX_CONNECTIONS.
* pkg/operator/controller/ingress/deployment_test.go
(TestDesiredRouterDeployment): Verify that desiredRouterDeployment sets
ROUTER_MAX_CONNECTIONS as expected.
* test/e2e/operator_test.go (TestMaxConnectionsUnsupportedConfigOverride):
Verify that the operator sets ROUTER_MAX_CONNECTIONS to the appropriate
value based on the value specified in the unsupported configuration
override if an override is specified.

When updating an ingresscontroller, do a get immediately before doing the
update.

This is a cleanup and should have no significant functional effect.

* test/e2e/operator_test.go (TestProxyProtocolAPI)
(TestLoadBalancingAlgorithmUnsupportedConfigOverride)
(TestDynamicConfigManagerUnsupportedConfigOverride)
(TestReloadIntervalUnsupportedConfigOverride): Get ingresscontroller
immediately before updating it.

Ignore not-found errors when deleting client pods.  These pods run and then
exit, and it is possible that they get garbage collected before the test
code deletes them, in which case it is expected that they be not found.

* test/e2e/canary_test.go (TestCanaryRoute):
* test/e2e/client_tls_test.go (TestClientTLS):
* test/e2e/http_header_buffer_test.go (TestHTTPHeaderBufferSize):
* test/e2e/http_header_name_case_adjustment_test.go
(TestHeaderNameCaseAdjustment):
* test/e2e/operator_test.go (TestHTTPHeaderCapture, TestHTTPCookieCapture)
(TestUniqueIdHeader): Ignore not-found errors when deleting client pods.