'nan' vs 'NaN' as default for CSV format

[youngsofun]

background:

#8913
#8914

[youngsofun]

we can limit it to only a few valid choices for now: "nan" or "NaN" or "null" or "NULL" ?

[youngsofun]

all so including "Nan"

[youngsofun]

https://note.nkmk.me/en/python-pandas-nan-none-na/

the value we use be accept by most csv lib in languages.

need to check this manually.

[youngsofun]

related #8941

[youngsofun]

maybe we can choose to unload NaN by default and can be configed with option.
and accept all variants in load, no need to config.

[ariesdevil]

I think that's the appropriate way to deal with data.

[youngsofun]

there is a little problem with “NULL” display
for column float NULL, can not tell apart Nan from NULL

[youngsofun]

and clickhouse TSV may require nan, we have to make clickhouse driver work.

need to check the code of clickhouse clients.