Autocomplete: sort by number of usages

[tleb]

Searching for clk_ gives:

clk_1000_1200_src
clk_1000_1440_src
clk_108MHz
clk_125
clk_12_24m
clk_12_24m_in
clk_12m
clk_12m_out_groups
clk_1440_1200_src
clk_168m

But those are almost useless. Instead, we could count the number of references for each symbols that matches the search. Code is here. That would raise to the top the subsystem defs that get used by many.

Some benchmarking would be required, but we don't see many requests on that endpoint so we have no reason to worry as long as the duration isn't anything outside the ordinary.

[fstachura]

4-character prefixes with largest number of hits have ~100k results on my instance (elixir.fbstc.org, random versions from v3.1 to v6.13), it may take a while to fetch values of all these keys and sort them

$ python3 dump.py | uniq -c | sort -n
 102278 b'DPCS'
 145081 b'C20_'
 169142 b'BIF_'

(dump.py just dumps 4-character prefixes of all keys in definitions.db)