Setup:

$ cat lst-pl_wrds.txt
łódka   dwa    trzy     cztery     piec
rok    jutro

NOTES:

• for demonstration purposes I’ve …
• added extra spacing between words
• added 2nd line of words

Modifying OP’s script to print some debugging info:

$ cat testme
#!/bin/bash

url="https://www.diki.pl/slownik-angielskiego?q="

for e in $(cat lst-pl_wrds.txt)
do
    typeset -p e

    for o in $e
    do
        url="${url}${o}&end"
        typeset -p url
#       firefox --new-tab "$url" & timeout 1 >nul
    done
done

Running the script generates:

$ ./testme
declare -- e="łódka"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&end"
declare -- e="dwa"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&enddwa&end"
declare -- e="trzy"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&enddwa&endtrzy&end"
declare -- e="cztery"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&enddwa&endtrzy&endcztery&end"
declare -- e="piec"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&enddwa&endtrzy&endcztery&endpiec&end"
declare -- e="rok"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&enddwa&endtrzy&endcztery&endpiec&endrok&end"
declare -- e="jutro"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&enddwa&endtrzy&endcztery&endpiec&endrok&endjutro&end"

Two things to notice from this output:

• 1) the e variable only ever contains a single string, but from the code (for o in $e) it appears OP is expecting multiple strings
• this happens because for e in $(cat lst-pl_wrds.txt) effectively becomes for e in łódka dwa trzy cztery piec rok jutro (all words placed on a single line with a single space as delimiter)
• in this particular case this behavior is ‘ok’ as long as we are dealing with items that contain spaces (eg, compound words, phrases, etc)
• which in turn means the inner loop is not required … in this case
• in the proposed answer (below) we’ll go ahead and keep the inner loop but we’ll need to modify the outer loop to work on a single line of input at a time
• 2) the url variable keeps appending each new word on the end of the previous url value
• this is occurring because that’s exactly what OP is requesting with url="${url}${o}&end"; OP is using the same url variable for the ‘base’ url as well as the ‘new’ url
• what OP probably wants is to use two variables, one for the base url and one for the word-related url

Modifying our current test script to address the above issues:

$ cat testme
#!/bin/bash

base_url="https://www.diki.pl/slownik-angielskiego?q="     # one variable for base url

while read -r -a arr                                       # split line on white space and place words in array "arr[]"
do
    typeset -p arr

    for o in "${arr[@]}"                                   # step through list of array items
    do
        url="${base_url}${o}&end"                          # different variable for word-related url
        typeset -p url
#       firefox --new-tab "${url}" & timeout 1 >nul
    done
done < lst-pl_wrds.txt

The modified script generates:

declare -a arr=([0]="łódka" [1]="dwa" [2]="trzy" [3]="cztery" [4]="piec")
declare -- url="https://www.diki.pl/slownik-angielskiego?q=łódka&end"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=dwa&end"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=trzy&end"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=cztery&end"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=piec&end"

declare -a arr=([0]="rok" [1]="jutro")
declare -- url="https://www.diki.pl/slownik-angielskiego?q=rok&end"
declare -- url="https://www.diki.pl/slownik-angielskiego?q=jutro&end"

From this output we see:

• our array ar[] is (re)populated with the words from a single line from our input file
• url is no longer appending multiple words together

Rolling these modifications into OP’s current script:

$ cat dictionary.sh
#!/bin/bash

base_url="https://www.diki.pl/slownik-angielskiego?q="

while read -r -a arr
do
    for o in "${arr[@]}"
    do
        url="${base_url}${o}&end"
        firefox --new-tab "${url}" >/dev/null &
        sleep 1
    done
done < lst-pl_wrds.txt

NOTES:

• I’m assuming OP’s use of >nul is meant to discard/ignore stdout from the firefox call; in this case the unix/linux equivalent is >/dev/null
• the link OP provides for firefox command line options does not include anything related to ‘timeouts’ so I’m assuming OP’s intention is to wait 1 second between each firefox invocation; if this assumption is wrong then OP can modify the firefox call as needed (and remove the sleep 1?)
  the & on the end of the firefox call will place the call in the background

Testing results:

• I use brave-browser so I replaced the firefox ... line with brave-browser "${url}" > /dev/null &
• when I run the script it opens new tabs at 1-second intervals in the active brave-browser window (ie, the brave-browser window I’m typing this answer into); each tab represents a single word and the tabs are displayed in the same order as they are read from the file:

NOTES:

• if OP finds the script has some issues displaying the last word from each line (piec and jutro in this example) then it may be an issue of the file containing windows/dos line endings (rn)
• OP can run od -c lst-pl_wrds.txt and if the output contains r entries then this confirms the windows/dos line endings exist in the file
• to remove the line endings OP can run dos2unix lst-pl_wrds.txt; this only needs to be run once as it will edit the file and remove the r characters