Large tables generation speedup

[cervinka]

Generating large tables (more than thousands of rows) takes a very long time.

I looked at the internal Sheet implementation and tried to rewrite it from a list of lists to a map of maps.

The result was a drastic speedup in generating large tables:

number of rows	current implementation	new implementation	speedup
10k rows	16s	0.7s	~23x
30k rows	154s	2.4s	~64x
50k rows	448s	2.7s	~166x
100k rows	1716s	5.7s	~301x
1M rows	didn't wait :-)	65.7s	?

(unscientific time measurement of generating tables with a given number of rows and 10 columns)

All my tests and use cases work with the new implementation, but it would definitely take more testing. Please don't take the pull request as final yet. I just want to make sure at this point that you would be interested in the modification and would consider including it in the main branch.

Anyone is of course welcome to test on generating their own data.

[xou]

Thank you so much, this looks really great.
I'd love to merge this, I intend to extend the test coverage in the upcoming weeks and will then have another look into merging this!