perf bug: repeated decompression of runend `ends`

[tpoterba]

What happened?

execute_scalar multiple times on a pco encoded array of ends:
find_slice_end_index in runend array:

root: vortex.runend(u16, len=524288) nbytes=13.97 kB (100.00%) 
  metadata: offset: 0                                          
  ends: vortex.pco(u32, len=27382) nbytes=8.29 kB (59.35%)     
    metadata: ptype: u32, nrows: 27382, slice: 0..27382        
    buffer: chunk_meta_0 host 76 B (align=1) (0.92%)           
    buffer: page_0 host 2.14 kB (align=1) (25.83%)             
    buffer: page_1 host 2.04 kB (align=1) (24.57%)             
    buffer: page_2 host 1.95 kB (align=1) (23.50%)             
    buffer: page_3 host 2.09 kB (align=1) (25.18%)             
  values: vortex.pco(u16, len=27382) nbytes=5.68 kB (40.65%)   
    metadata: ptype: u16, nrows: 27382, slice: 0..27382        
    buffer: chunk_meta_0 host 19 B (align=1) (0.33%)           
    buffer: page_0 host 1.45 kB (align=1) (25.48%)             
    buffer: page_1 host 1.38 kB (align=1) (24.34%)             
    buffer: page_2 host 1.38 kB (align=1) (24.34%)             
    buffer: page_3 host 1.45 kB (align=1) (25.50%)             

Steps to reproduce

See above

Environment

exists in latest

Additional context

No response