feat: Cache accessed source files

[mark-koch]

We used to store the entire source code that was used to build an AST as a field within every node of the AST. This was a hack so that we can render the source line from an AST node when an error occurs. With our new diagnostics system, we should do better than that!

With this PR, we cache the source of every file accessed by the compiler in a new SourceMap class. Then, we can look up the source associated with a span when rendering a diagnostic.

Note that we can't rely on Pythons linecache since we also need to keep track of cells in jupyter notebooks as individual "files".

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 91.48936% with 4 lines in your changes missing coverage. Please review.

Project coverage is 90.22%. Comparing base (d671bb5) to head (40e8cf2).

Files with missing lines	Patch %	Lines
guppylang/definition/function.py	75.00%	2 Missing ⚠️
guppylang/span.py	83.33%	2 Missing ⚠️

Additional details and impacted files

@@                 Coverage Diff                  @@
##           feat/diagnostics     #551      +/-   ##
====================================================
+ Coverage             89.90%   90.22%   +0.32%     
====================================================
  Files                    61       61              
  Lines                  6347     6376      +29     
====================================================
+ Hits                   5706     5753      +47     
+ Misses                  641      623      -18     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[doug-q]

Looks good. I think the various parse methods that take and never use a SourceMap are various constructs that don't track their source location yet?

[mark-koch]

I think the various parse methods that take and never use a SourceMap are various constructs that don't track their source location yet?

Yes, exactly!