Rename generic operator.

[var77]

Rename generic operator <-> to <?> and fix update script for 0.0.9->0.0.10

---

There was an issue in update script: the tags were being sorted in asc order for ascii codes so the to_tags were like ['0.0.10', '0.0.4', '0.0.5', '0.0.6' ...] while we expect the latest tag to be the last element. So because of it the to_tag to be tested was being selected as 0.0.9 and there was a case when the update script was being run as from_tag=0.0.9 to_tag=0.0.9 and it made database to crash from this assert assert(hdr->blockmap_groups_nr == groupno + 1); in external_index.c:236

[codecov]

Codecov Report

Merging #244 (8eb6023) into main (896ed5a) will decrease coverage by 0.34%.
The diff coverage is 60.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #244      +/-   ##
==========================================
- Coverage   78.06%   77.72%   -0.34%     
==========================================
  Files          23       23              
  Lines        1805     1791      -14     
  Branches      455      455              
==========================================
- Hits         1409     1392      -17     
- Misses        201      206       +5     
+ Partials      195      193       -2     

Files	Coverage Δ
src/hnsw/options.c	82.53% <100.00%> (+1.58%)	⬆️
src/hooks/utils.c	81.81% <100.00%> (ø)
src/hnsw.c	79.90% <50.00%> (-0.36%)	⬇️

... and 1 file with indirect coverage changes

[github-actions]

Benchmarks

metric	old	new	pct change
recall (after create)	0.740	0.740	-
recall (after insert)	0.732	0.778	+6.28%
select bulk tps	481.860	478.706	-0.65%
select bulk latency (ms)	15.515	15.834	+2.06%
select bulk latency (stddev ms)	3.669	4.055	+10.52%
create latency (ms)	1186.241	1205.942	+1.66%
insert bulk tps	10.948	11.169	+2.02%
insert bulk latency (ms)	91.328	89.525	-1.97%
insert bulk latency (stddev ms)	3.173	2.897	-8.70%
disk usage (bytes)	6348800.000	6348800.000	-

[Ngalstyan4]

Looks good!

[Ngalstyan4]

should this be USING lantern_hnsw?

[Ngalstyan4]

or use am_name? It is the code I had written in the last PR and I think I forgot to consider the changed naming of our AM when pgvector is present

[var77]

True, that should be lantern_hnsw;

[Ngalstyan4]

hm, do you know why before the Sort Key function was schema-qualified (public.l2sq_dist) while now it is not (l2sq_dist) ?

[var77]

Actually that was added after the latest release, before that it was without public if I remember correctly, not sure why but one change that I did to that operator was adding it for integer type as well