Allow encoding data with missing fields.

[scossu]

This small PR allows encoding CSV from data entirely missing a field that was indicated in the fieldsToKeep option.

Useful when encoding multiple data sets into similarly formatted CSV files.

>  csv = require "ftcsv"
>  data = {
>> {a = 1, b = 2, c = 3},
>> {a = 10, b = 20},
>> {a = 100, c = 200},
>> }
>  csv.encode(data, {fieldsToKeep = {"a", "b", "c", "d"}})
/home/ste/.luarocks/share/lua/5.4/ftcsv.lua:757: ftcsv: the field 'd' doesn't exist in the inputTable

Stack trace:
	#0 [C]: in function ?
	#1 [C]: in function 'error'
	#2 /home/ste/.luarocks/share/lua/5.4/ftcsv.lua:757: in function 'validateHeaders'
	#3 /home/ste/.luarocks/share/lua/5.4/ftcsv.lua:817: in function 'initializeGenerator'
	... tail calls
	#4 /home/ste/.luarocks/share/lua/5.4/ftcsv.lua:827: in function ?
	#5 [C]: in function 'enter'
	#6 ...luarocks/lib/luarocks/rocks-5.4/luaprompt/0.9-1/bin/luap:230: in the main chunk
	#7 [C]: in function ?

>  csv.encode(data, {fieldsToKeep = {"a", "b", "c", "d"}, allowEmpty = true})
_[2] = [[
"a","b","c","d"
"1","2","3","nil"
"10","20","nil","nil"
"100","nil","200","nil"
]]
>  

[FourierTransformer]

Ahhh yeah, that makes sense and looks good overall! I think maybe it should be changed to allowMissingKeys instead of allowEmpty to try and make it less ambiguous (open to other suggestions as well).

Oh and maybe add a few unit tests. I'm assuming it also works with the encodeNilAs option?

[scossu]

In my implementation it works with EncodeNilAs. I agree on renaming. I will do that and add some tests in the next day or two. Thanks.

[scossu]

I added a unit test; it only runs on a dedicated CSV/JSON pair because otherwise it breaks simple_crlf. Other tests are failing as well, but they don't seem related to this PR.

[FourierTransformer]

All the tests pass on the main branch: https://github.com/FourierTransformer/ftcsv/actions/runs/17284534932/job/49059071095. I have no intention of changing existing behaviour, and no plans for a major release, so at the moment there can't be any breaking changes to exisitn tests. You may want to look into what's causing the failures. Also, I think you may have accidentally included other new options noHeader and noData.

[scossu]

Oops, I'm sorry, those were options I was experimenting with and that weren't completed. I will remove them.

[scossu]

Tests pass now. Is anything else missing for approving this PR?