improve latency: specify element type explicitly in each comprehension

[nsajko]

This improves latency a bit.

[nsajko]

Script for timing test_all on an empty package:

using Test: @testset
using Aqua: Aqua
using Brick: Brick  # empty package
@timev @testset "test set" begin
    Aqua.test_all(Brick)
end

Experiment is run like so:

• with Julia v1.11.2
• with all packages precompiled beforehand
• in a noninteractive session, by specifying the script file name on the command-line

Results on Git master branch:

Test Summary: | Pass  Total  Time
test set      |   11     11  7.4s
  7.580602 seconds (5.57 M allocations: 319.143 MiB, 1.10% gc time, 37.78% compilation time)
elapsed time (ns):  7.580601886e9
gc time (ns):       83028137
bytes allocated:    334646112
pool allocs:        5562994
non-pool GC allocs: 1567
malloc() calls:     3955
free() calls:       3819
minor collections:  5
full collections:   0

Results on PR branch:

Test Summary: | Pass  Total  Time
test set      |   11     11  7.1s
  7.238491 seconds (5.09 M allocations: 295.169 MiB, 1.33% gc time, 35.07% compilation time)
elapsed time (ns):  7.238490548e9
gc time (ns):       96283002
bytes allocated:    309506936
pool allocs:        5082786
non-pool GC allocs: 1546
malloc() calls:     3561
free() calls:       2783
minor collections:  3
full collections:   1

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 74.90%. Comparing base (1fca5d9) to head (7d9293b).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #314   +/-   ##
=======================================
  Coverage   74.90%   74.90%           
=======================================
  Files          11       11           
  Lines         761      761           
=======================================
  Hits          570      570           
  Misses        191      191           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[nsajko]

Am I supposed to honor the automatic code style suggestions? Seems like that would unnecessarily increase diff size of the PR.