Note: this is a very simple approach. Brightness is a well discussed subject with a lot of options. For sophisticated solutions you often also drag in saturation and much more. Per pixel options are maybe not the best approach, but for the sake of this post I have constructed an answer that will solve your specific problem below.
// edit 2
Thinking about this some more, I did not think about the solution to the equation not being unique. You indeed need to store the original and recalculate from the original image. I would still advice using an approved brightness equation like the ones found in the link above. Simply modifying R,G, and B channels might not be what your users expect.
The below answer must be combined with working on the original image and displaying a modified copy as mentioned in other answers.
I would not increase R, G, and B channels directly but go with a perceived brightness option like found in here.
Lets say you take:
L = (0.299*R + 0.587*G + 0.114*B)
You know the min(L) will be 0, and the max(L) will be 255. This is where your numeric up/down will be limited to [0,255]. Next you simply increase/decrease L and calculate the RGB using the formula.
//edit
You case as example:
r=100 , g = 200 ; b =125
L = (0.299*100 + 0.587*200 + 0.114*125)
L = 161.5
Now lets go to the max (limited) to get the extreme case and see this still works:
L = 255
L = (0.299*255 + 0.587 * 255 + 0.114 * 255)
RGB = (255,255,255)
Going back will also always work, 0 gives black everything in between has a guaranteed RGB in range R,G,B in [0,255].