made number field search result download more efficient

[JohnCremona]

See Issue #1079. It is now possible to download 1.36 million number fields in a little over 2 minutes, no timeout. The key is to avoid constructing the WebNumberField object for each.

[JohnCremona]

If you like this we can also think about doing what Drew and I did for the equivalent genus 2 curves function: more slick switching between language options, and also defined a function make_data() in each language so that (for example in Magma) after the user reads in the file they type curves:=make_data();
and then curves is a Magma list of genus 2 curves. This saves the user from having to work out how to create the objects from what we give them and would often be useful.

[jwj61]

This produces the downloads faster. It seems stylistically wrong to have bits that need to know how items are stored in the database outside webnumberfield, although I can see how to work around that. This version does not handle the case where the class number/group are not computed. Again, that is not hard to fix.

But, I don't think this solves the timeout issue. I believe beta and prod kill web requests which take more than 30 seconds, and 2 minutes is much more than that. While it works when running your own copy, I think the same download will be killed and end with a proxy error if pushed to beta/prod.

In the next day or two I will put in a pull request doing what I allude to in the first paragraph, and remove the download links when the number of results is more than 100000.

[JohnCremona]

You may be right. The main thing I was looking for was whether WebNF created a Sage NF. If there are slow parts of the WebNF to create they can perhaps only be done when actually needed so that here we could have a "WebNF lite".

[jwj61]

I don't think webnumberfield was should not be doing anything aside from creating a copy of the data (because it was using the from_data construct). I see now creating code snippets was added to the init routine, which means a hard disk access to initialize. That is a waste of time here and should be done only when actually needed.

[JohnCremona]

This was superseded by #1091 so I'm closing it.