Can I still use a LMER with the following Shapiro-Wilk normality test result on the residuals?

The assumption of normality seems to show it's not normal, p=0.02 (142 observations) Though I think the QQ graph (below) may be not too bad?

if so, how would I justify continuing with this? just say that the QQ graph shows it's close?

If using LMER is not valid, what other tests could perhaps be used?

All help is greatly appreciated! thanks!

---- Model

m <- lmer(Amount ~ Diet + Side + (1|PairID), data = dataDietChoice)

• amount is the quantity eaten (trying to work out if there is an effect of side or diet)
• side is left/right (there about 10 feeder pairs, sides containing each diet were switched daily),
• diet is 'high nutrient' or 'low nutrient' (the model originally also had Day but this was not significant and was removed)
• pairID is the pair of feeders, there were about 10 pairs

____ R results

> Shapiro-Wilk normality test
> 
> data:  residuals_m W = 0.978, p-value = 0.022
 


    > print(levene_test_diet)
    Levene's Test for Homogeneity of Variance (center = median)
    
     Df     F value  Pr(>F)
    
    group   1 0.69 0.40725
    
          140                             
    
    > print(levene_test_side)
    Levene's Test for Homogeneity of Variance (center = median)
    
           Df   F value Pr(>F)
    group   1     1.11    0.29
          140        

sorry if this has already been asked, just can't find (or perhaps understand) the other answers. I have seen an answer that says based on the QQ graph it is ok, I'm not sure how to know how close to the line the QQ graph needs to be for that to be valid.

I'm happy to change the test if another test is better.