Simplest attempt at correcting issue #1122.
When movescu is called with +B (--bit-preserving) option, the output file ends up in the current working directory as opposed to the desired output directory even if you add the -od (--output-directory) option in the command line.

Note that a better approach could be to have a singular writing function and adjust the function signatures to pass whether other processing logic should be skipped. However, that could cause a lot of headaches ensuring all logic paths are checked against the current bit-preserving behavior of outputting the data as-is.

I leave the option to the maintainer if to request further work to address this issue more comprehensively or if the change proposed here is good enough for the project.

The changes here appears to work as intended in my test environment.

Stylistic correction to comments

Unlike movescu, storescp does not recompute the file path in the storeSCP callback. It assumes that imageFileName already includes the directory name. However, I separated the imageFileName generation from the full path generation, so I realized I needed to correct the deleteFile logic as well. Will need to do something like this in movescu to ensure consistency across both codebases.

Keeping consistency between movescu and storescp so they both look at the file path in a similar manner.

Attempting to resolve merge conflict that arose in storescp due to new commits in the upstream mirror. It looks like it is addressing the   path construction with a slightly different approach. A second commit will be added after I check if the directory path is getting added twice.

Now that master is using snprintf(), I decided to remove my previously proposed solution in favor of the snprintf() suggestion. Also, adding bracket to if statement per comments in PR.