You can configure the maximum indexable size of an attachment - https://confluence.atlassian.com/conf57/configuring-attachment-size-701435663.html

However, the way this works is that the content is still all extracted and then simply not saved in the index.

Customers often use this setting because large documents are resource intensive to extract the content from and often fail. With the current implementation we're helping to not store the content, but we're still expending resources to extract the content.

Recommend that this setting be modified so that these files are excluded earlier in the process and not extracted at all.