"Last 24 hours" in "Indexer Failures" on overview page should refer to index time, not message timestamp.

## Expected Behavior

If an indexing error happens within the last 24 hours, the `Indexing Errors` section on the System/Overview page should turn red, regardless of the timestamp in the message.

The 24 hours period should refer to the time the error happened, not the timestamp in the message. If I'm ingesting old logs, I still want to be informed of indexing errors.

## Current Behavior

If an indexing error happens when ingesting a message, but the message has a timestamp older than 24 hours, the "Indexer Failures" section on the overviews page stays green.

## Possible Solution

Use the timestamp of the indexing attempt when storing the failures, not the message timestamp.

## Steps to Reproduce (for bugs)

1. Send a message with a timestamp and a field set to a numeric value
2. Send a message with a timestamp older than 24 hours and the field set to a string value.
3. Observe that indexing fails but the "Indexer Failures" section stays green.

* Graylog Version: `5.0.0-beta.4+2d0b0a6`