Impute TIFF xres/yres from withMetadata({density})

[mbklein]

Closes #2941

This PR uses the density value passed to withMetadata() as a fallback if the output format is TIFF and netter xres nor yres is provided in the TIFF options. I've also updated the API documentation with to include a short explanation of this behavior.

[mbklein]

@lovell While creating two new tests for this functionality, I discovered that I couldn't get them to fail at all when written as promises. I believe this is due to the fact that the it() function completes (and passes) before the promise resolves, as described in this guide. I tried using the done() callback, but found I could achieve better and more consistent results by using asynchronous test functions with async/await. I'm pretty sure this same issue might cause some false passes in other parts of the test suite. I would be happy to refactor the rest of asynchronous code tests to use async/await as well, but that would obviously be the subject of a different pull request.

[lovell]

Hi Michael, thank you very much for this PR, and always a pleasure to see unit tests too.

I wonder if it might be a little simpler to have this logic in the C++ layer, perhaps something like the following:

--- a/src/pipeline.cc
+++ b/src/pipeline.cc
@@ -1496,6 +1496,9 @@ Napi::Value pipeline(const Napi::CallbackInfo& info) {
   baton->tiffTileHeight = sharp::AttrAsUint32(options, "tiffTileHeight");
   baton->tiffXres = sharp::AttrAsDouble(options, "tiffXres");
   baton->tiffYres = sharp::AttrAsDouble(options, "tiffYres");
+  if (baton->tiffXres == 1.0 && baton->tiffYres == 1.0 && baton->withMetadataDensity > 0) {
+    baton->tiffXres = baton->tiffYres = baton->withMetadataDensity / 25.4;
+  }
   // tiff compression options
   baton->tiffCompression = static_cast<VipsForeignTiffCompression>(
   vips_enum_from_nick(nullptr, VIPS_TYPE_FOREIGN_TIFF_COMPRESSION,

I think this should still satisfy the test cases you added.

[mbklein]

I skimmed pipeline.cc for an appropriate spot to add it, but came up short. This is much more concise. I'll make the change and force-push my branch.

[coveralls]

Coverage remained the same at 100.0% when pulling 33b5da3 on mbklein:tiff-density into b33231d on lovell:master.

[lovell]

Brilliant, thank you, landed via commit 342de36