Support source maps between editor and debugger file paths

[gpanders]

When nvim-dap is running on a separate machine from the debugger then Nvim and the debugger will use different paths for source files. For example, if Nvim is running on a machine where the source tree is located at /a/project/foo and the debugger is running on a separate machine where the source tree is located at /b/project/foo, when nvim-dap tries to set a breakpoint it will use /a/project/foo/ in the breakpoint location. When the debugger receives this request it will not set the breakpoint correctly because the debugger does not have any files at /a/project/foo.

The source_map option instructs nvim-dap to convert paths as seen by the editor and nvim-dap itself to the paths as the debugger sees them. The opposite mapping takes place when nvim-dap receives stack frame or source file information from the debugger.

Note that this is different than the "source map" option provided by many debuggers themselves (such as codelldb's "sourceMap" option or GDB's "substitute-path" setting). Those options convert compiled source paths (i.e. the path of the source file when the binary was compiled) into the debugger's path namespace (/b/project/foo in our example).

[brettmitchelldev]

This is fantastic. I routinely use volume mounts inside containers to develop, and it's hacky and inconvenient to mirror the host's directory structure inside the container.

I've tried this branch out and it solves the problem really well. 🚀

Perhaps the configuration could be assigned dynamically on a per-config and/or per-attach basis? Might make automation easier. Right now I'm automating assignment by mutating adapter.options.source_map directly by reference on attach.