Loading multiple .csv files uses ~double the memory it should

[krynju]

I'm on Julia master/1.8
and Windows

I tried CSV.File, CSV.read with NamedTuples, DTable, DataFrame etc. all have the same issue.
There's just some sticky memory leftover from loading the csv files

The table generated is about 1.6GB

# generate
using DataFrames
d = DataFrame((;[Symbol("a$i") => rand(Int32(1):Int32(1000), Int(1e8)) for i in 1:4]...));
# run GC.gc() a few times, memory usage settles at ~

# prep
using CSV
genchunk = () -> (; [Symbol("a$i") => rand(Int32(1):Int32(1000), Int(1e7)) for i = 1:4]...)
mkpath("data")

for i = 1:10
    CSV.write(joinpath(["data", "datapart_$i.csv"]), genchunk())
end

# load from multiple files
d = CSV.read(files, DataFrame)

generated

loaded from files

[krynju]

related to:
#850 (comment)
JuliaLang/julia#42566 (comment)

[krynju]

d = CSV.read(files, DataFrame, types=Int32)
forgot it parses as Int64 and that's where my double memory usage was coming from
this sticky memory related to glibc issue is still observable on my end though, but that's a different issue