improve long data performance

[amtoine]

cc/ @kubouch

in this PR, the rendering of data with rows will only compute the rendered strings for the rows in the frame.

from what i've seen, it looks like it allows to run nu_plugin_explore on data longer by around an order of magnitude 🤔

• before this PR
  • seq 1 100_000 | nu_plugin_explore is already pretty slow (slower than with seq 1 1_000_000 below)
  • seq 1 1_000_000 | nu_plugin_explore is absolutely unusable
• with this PR
  • seq 1 100_000 | nu_plugin_explore is very responsive
  • seq 1 1_000_000 | nu_plugin_explore starts to feel a bit slow
  • seq 1 10_000_000 | nu_plugin_explore starts to be very slow, sometimes it does not even respond and nu_plugin_explore took ages to start (i'm wondering if the plugin has to collect before starting, and collecting $10^{7}$ items might be pretty slow)

bonus

this PR also fixes the height of the frame in the compact layout, making it consistent with the table layout and the rendering style of true tables.

[amtoine]

would be great to have some benchmarks for this