Proposal: ScriptV2 bundler

[macobo]

Changes

This PR proposes using rollup as a bundler for tracker script.

Benefit:

• Splits the code across multiple files

Drawbacks:

• More dependencies
• It increases the legacy script size due to needing to use config.domain/config.endpoint/config.autoCapturePageviews over vars and new function definitions - this doesn't minify as nicely. See comment below for stats.

Best reviewed commit-by-commit.

[github-actions]

Analyzed 1025 tracker script variants for size changes.
The following tables summarize the results, with comparison with the baseline version in parentheses.

Main variants:

	Brotli	Gzip	Uncompressed
plausible-web.js	2281B (-1B / 0%)	2601B (-10B / -0.4%)	6123B (+93B / +1.5%)

Important legacy variants:

	Brotli	Gzip	Uncompressed
plausible.js	1140B (+24B / +2.2%)	1329B (+31B / +2.4%)	2883B (+152B / +5.6%)
plausible.hash.js	1103B (+19B / +1.8%)	1294B (+26B / +2.1%)	2769B (+146B / +5.6%)
plausible.pageview-props.tagged-events.js	1842B (+16B / +0.9%)	2145B (+35B / +1.7%)	4797B (+134B / +2.9%)
plausible.file-downloads.hash.pageview-props.revenue.js	1657B (+19B / +1.2%)	1954B (+26B / +1.3%)	4139B (+134B / +3.3%)
plausible.compat.exclusions.file-downloads.outbound-links.pageview-props.revenue.tagged-events.js	2248B (+24B / +1.1%)	2656B (+37B / +1.4%)	5937B (+138B / +2.4%)

Summary:

	Brotli	Gzip	Uncompressed
Largest variant (plausible.compat.exclusions.file-downloads.outbound-links.pageview-props.revenue.tagged-events.js)	2248B (+24B / +1.1%)	2656B (+37B / +1.4%)	5937B (+138B / +2.4%)
Max change (plausible.compat.local.revenue.js)	1040B (+36B / +3.6%)	1246B (+35B / +2.9%)	2690B (+152B / +6%)
Min change (plausible.exclusions.hash.local.manual.revenue.js)	1052B (-3B / -0.3%)	1277B (+19B / +1.5%)	2547B (+126B / +5.2%)
Median change	1696B (+21B / +1.2%)	2006B (+28B / +1.4%)	4316B (+128B / +3%)

In total, 1014 variants brotli size increased and 8 variants brotli size decreased.

[apata]

Fleshing out the benefits and drawbacks more, I see the following

• probably increased development ease from better code organization
• maybe increased risk of invalid code from needing to consider module to module interactions
  • (can be mitigated with linter, maybe typescript)
• increased supply chain attack risk from additional binary
• script size increase

@macobo Did you already try organizing the code in the single file? If you did and found it still hard to comprehend, then I think the drawbacks, small as they are, are acceptable. Even if you didn't try that, I'd be ok with us proceeding with rollup for sake of development velocity.

[apata]

After going through the changes, it is easier to understand now, especially engagement. Good job!

I'm worried that I might have missed some accidental changes during the refactor though, for features that the tests don't cover. Third set of eyes on this maybe?

Edit: Ah, I see that there was already a third set of eyes on it, good!