Add variant_length to output for plink and tskit, make missing arrays for index an error

[benjeffery]

Fixes #379

[coveralls]

coverage: 98.179% (+0.01%) from 98.166%
when pulling 7049588 on benjeffery:variant_length
into 7942838 on sgkit-dev:main.

[jeromekelleher]

LGTM - we can probably refine the API a bit, but good step in the right direction.

@tomwhite does this look right to you in terms of computing variant length? Should we do this, storing a "length" field everywhere, or have some fallback in the index computation code which computes the default when it's missing?

[tomwhite]

@tomwhite does this look right to you in terms of computing variant length? Should we do this, storing a "length" field everywhere, or have some fallback in the index computation code which computes the default when it's missing?

This looks good to me. It's not obvious to me that there is a general way of computing it so having the formats provide is safest.

[tomwhite]

Suggested change

	# Only three of the fixed fields have a direct one-to-one mapping.
	# Only two of the fixed fields have a direct one-to-one mapping.

[benjeffery]

Thanks, fixed

[benjeffery]

Rebased and ready to go

[jeromekelleher]

Looks like there's a real coverage drop here, with some warnings that used to be tested no longer covered. We should address these (in general, coverage drops aren't something we want)

[benjeffery]

It's the index-when-arrays-missing error case has been uncovered (it was accidentally covered before). I'll add a specific test for it.

[benjeffery]

Test added and rebased. Coveralls now happy.