Large lookup tables (CSV file adapter) can crash Graylog node

## Expected Behavior

When creating a lookup adapter (example: CSV file adapter), it shouldn't exhaust the heap memory of the Graylog node, even if its backing contents don't fit into memory.

## Current Behavior

Creating a CSV file adapter backed by a large file (example: 250 MB) will ultimately exhaust the (heap) memory of the Graylog node if the file is larger than the available memory (or even smaller because the resident memory size is a multiple of the size on disk).

Refs https://community.graylog.org/t/graylog-2-4-3-crash-from-lookup-table/3942

## Possible Solution

Either issue a warning and deny creating a lookup adapter which is backed by contents not fitting into the available heap memory or, in case of the CSV file adapter, don't pre-load everything into memory from disk but use some paging mechanism instead.

## Steps to Reproduce (for bugs)

1. Start Graylog with 2 GB of heap memory
2. Create 250 MB large CSV file with lots of entries
3. Create lookup table backed by CSV file adapter configured to read the previously created file
4. ❓
5. Profit! 💸

## Your Environment

* Graylog Version: 2.4.x
* Elasticsearch Version: N/A
* MongoDB Version: N/A