Fix relative path loading of plugins

[cmdcolin]

Fixes #3262

It adds two examples of loading a UMD plugin to the volvox config

Firstly, this PR makes umdUrl, an existing method for configuring plugin URLs, always loaded relative to index.html. The issue #3262 describes why this wasn't working with relative paths currently

Secondly, this PR also adds new settings, esmLoc/umdLoc, which resolve the plugin location relative to the current config.json being used, which is similar to other "loc" objects in our config files (e.g. track data file locations like bam, bigwig, gff, etc are assumed to be in the same "folder" as the config.json)

I didn't add cjsLoc (commonjs) yet, used for desktop plugins, but can add it if there is interest

Finally, I added examples to the volvox config using umdUrl and umdLoc. I did NOT use esmUrl/esmLoc because it does not work in firefox with webworkers. firefox cannot load ESM modules in workers currently, though they may add it in the future.

As a result, I added a "minimal" UMD style plugin that can be handwritten. It's more annoying to write this way than ESM but, more browser compatibility

Example minimal plugin

class UMDLocPlugin {
  name = 'UMDLocPlugin'
  install(pluginManager) {}
  configure() {}
}

// the plugin will be included in both the main thread and web worker, so
// install plugin to either window or self (webworker global scope)
;(typeof self !== 'undefined' ? self : window).JBrowsePluginUMDLocPlugin = {
  default: UMDLocPlugin,
}

[cmdcolin]

also, I removed the strict check for the plugin class being an instanceof Plugin, since i think it is easier to just say that it is a basic class. If needed, we can make the handwritten plugin "extend JBrowseExports["@jbrowse/core/Plugin"]" or something like that to fulfill instanceof but I think it may not truly be needed

[cmdcolin]

this was unused :)

[cmdcolin]

i didn't add checks for e.g. umdLoc here because since i didn't think our plugin store would use this

[codecov]

Codecov Report

Merging #3266 (d1545a8) into main (4018477) will increase coverage by 0.04%.
The diff coverage is 21.56%.

@@            Coverage Diff             @@
##             main    #3266      +/-   ##
==========================================
+ Coverage   59.45%   59.49%   +0.04%     
==========================================
  Files         676      676              
  Lines       28828    28807      -21     
  Branches     7039     7041       +2     
==========================================
+ Hits        17139    17140       +1     
+ Misses      11415    11393      -22     
  Partials      274      274              

Impacted Files	Coverage Δ
packages/core/rpc/RpcManager.ts	70.58% <ø> (ø)
packages/core/rpc/WebWorkerRpcDriver.ts	0.00% <0.00%> (ø)
products/jbrowse-desktop/src/StartScreen/util.tsx	0.00% <0.00%> (ø)
products/jbrowse-desktop/src/rpc.worker.ts	0.00% <0.00%> (ø)
products/jbrowse-web/src/rpc.worker.ts	0.00% <0.00%> (ø)
packages/core/PluginLoader.ts	26.43% <31.70%> (+2.26%)	⬆️
products/jbrowse-web/src/SessionLoader.ts	63.92% <40.00%> (+2.62%)	⬆️
packages/core/PluginManager.ts	90.58% <50.00%> (-0.65%)	⬇️
...rative-view/src/ServerSideRenderedBlockContent.tsx	64.00% <0.00%> (-4.00%)	⬇️
products/jbrowse-web/src/util.ts	48.33% <0.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[cmdcolin]

@carolinebridge-oicr I updated the no-build plugin tutorial to reflect some of the lessons learned from this PR, and tagged you for review too.

I changed the no-build tutorial to use a UMD type plugin, and removed the unpkg code since that would use ESM. Perhaps could be added back in the future if ESM gets supported by firefox

[cmdcolin]

I think this should be good to go...might go ahead with merge

[garrettjstevens]

also, I removed the strict check for the plugin class being an instanceof Plugin, since i think it is easier to just say that it is a basic class. If needed, we can make the handwritten plugin "extend JBrowseExports["@jbrowse/core/Plugin"]" or something like that to fulfill instanceof but I think it may not truly be needed

Might be good to at least check that the plugin has install and configure methods, then. Other than that, looks good!

[cmdcolin]

cool :) added checks for install and configure at least existing as properties

made it so that 'install' and 'configure' substitute for the instanceof-style check d1545a8 (had done a different thing to detect the different types of 'load' records, but this seems better)