Drop crates.textsearchable_index_col from the export

[jtgeibel]

This column is postgres specific and is normally populated via a
trigger. The trigger is now enabled during the import so that the column
can be dropped from the export.

This addresses part of what was raised in bullet point 2 of #2078. The large readme column remains because there could be people using that data, but the text search column is redundant.

r? @smarnach
cc @kornelski

[pietroalbini]

The code changes look good!

I tested locally what the impact of this would be for how long it takes to import a database dump, since the trigger would have to be executed every crate we insert. To test that I downloaded today's dump, removed version_downloads as that takes ages and is not relevant and I made a copy of the dump with the changes of this PR (enabling the trigger and removing the column from the dump):

• Status quo: 42.3 seconds (176 MB crates.csv)
• With this PR: 56.7 seconds (95 MB crates.csv)

So, with this importing the dump is 34% slower. We're removing 46% of the data in crates.csv though, which will definitely help download size and everyone who parses the csvs manually instead of importing them in PostgreSQL.

Personally I'd say the tradeoff is worth it.

[jtgeibel]

Thanks for the analysis of the space/time tradeoff. I've pushed a new commit with your consistency suggestion.

[pietroalbini]

@bors r+

[bors]

📌 Commit 8fb7d30 has been approved by pietroalbini

[bors]

⌛ Testing commit 8fb7d30 with merge e421caa...

[bors]

☀️ Test successful - checks-actions
Approved by: pietroalbini
Pushing e421caa to master...