Speedup & cosmetics

[tophf]

• 1000x speedup for 200k tracked lines - just by reducing the progress update frequency to 1upd/sec.
• Use Aegisub 3.0 subtitle API (delete & insert with multiple values) if possible
• Insert new lines after the last originally selected line instead of the first one - the new implementation is assuming the original selection was contiguous, which is quite probable in fact due to the only custom sort method available is 'by time', meaning the user explicitly/obsessively wants the lines to be grouped in contiguous blocks per each frame/framespan, so we can easily expect he/she has arranged the original lines in a contiguous & sorted by time block as well.
• Fix accel char for [x] Shadow dialog checkbox.

[tophf]

I should talk to plorkyeran about a version that takes an array of indices rather than a list of arguments.

So now subs.delete is added which is good but:

1. no _subs.insert index, table_oflines yet
2. to support the existing Aegisub 3.x versions you'll have to wrap it in pcall as well as keep the existing code

    ok, _ = pcall sub.delete sel
    ok, _ = pcall -> sub.delete unpack sel unless ok
    if not ok --pay the price
        for i = #sel, 1, -1
            sub.delete sel[i] -- BALEET (in reverse because they are not necessarily contiguous)
            check_user_cancelled!

    ok, _ = pcall -> sub.insert strt, unpack [v.data for v in *lines]
    if not ok --pay the price
        for i, v in ipairs lines
            aegisub.progress.set i/#lines*100
            sub.insert strt, v.data -- not sure this is the best place to do this but owell
            strt += 1
            check_user_cancelled!

[torque]

I didn't even notice subs.insert(start, line1, line2, ...) existed. I wonder if that would be more efficient than inserting 1-by-1, though, because I imagine Lua has to copy memory around every time you insert an index into a table, which would get expensive with a big table full of dialog line tables.

That said, it would make sense for symmetric syntax to exist, and I could be wrong about it being slower, anyway.

[tophf]

You could be right. But we can try to reorganize processing of the 'lines' table like this (when and if subs.insert is extended):

    sortF = ({
        Time:   (l, n) ->
            l.key, l.num = l.start_time, n
            l
        ...................
    })[sor]
    ...................
    ok, _ = pcall sub.insert strt, lines
    if not ok --pay the price
        for i, v in ipairs lines
            aegisub.progress.set i/#lines*100
            sub.insert strt, v
            strt += 1
            check_user_cancelled!