Retention values that are "too big" overflow and cause confusing error messages

__Steps to reproduce:__

1. Attempt to create a bucket with retention.EverySeconds = 999999

__Expected behavior:__
Either the request succeeds, or a helpful error explaining the problem is returned

__Actual behavior:__
```
internal error: retention policy duration must be at least 1h0m0s
```
Debugging showed me that the retention value is overflowing in [this logic](https://github.com/influxdata/influxdb/blob/master/tenant/http_server_bucket.go#L365-L369).

The same bug is also present on bucket-update requests. See the original community report [here](https://community.influxdata.com/t/expiration-seconds-error-occurs-when-new-retention-time-is-99999-days/20464).

Start in tenant/http_server_bucket.go around lines 365-369 and reproduce the overflow with retention.EverySeconds set to 999999. Check both bucket creation and update requests, then verify that the request succeeds or returns a helpful error instead of the confusing minimum-duration message.

Written by the indexing model from the issue text.