Allow clicking on lower Manhattan variants

[pjvandehaar]

Unfortunately, bins are 3Mb wide, but LZ doesn't currently handle more than 400kb.

• Option 1: (see On Manhattan, improve peak selection and keep 500 variants union 500 peaks unbinned. #68) Improve peak selection to catch up to 1000 peaks better than 1e-6, and make all peaks clickable.

• Option 2: (see Handle genes that are longer than max-LZ-region #51) Make LZ handle 3Mb regions by subsetting variants, and link bins to LZ. Either (a) make pheno1_pval_le_1e-2.tsv.gz for each pheno or (b) use a priorityqueue to get best 1000 variants for each LZ request. Either way grey-out the missing rectangle of the plot. Somehow make genes panel denser or prioritize genes (protein-coding, etc).

• Option 3: record the position of the strongest pval for each bin, and link to that. (Nope.)