ref(build): Remove constToVarPlugin

[lforst]

This PR started out as an experiment but thinking more about this, I would actually go forward with this change, because:

• It fixes cases where replacing const with var causes actual problems (ui.vue application render is always cancelled #5917)
• Bundle size impact is like +23 bytes gzipped for the React SDK, +90 bytes gzipped for the Next.js SDK which is negligible, and it even reduces gzipped bundle size for the CDN bundles (???)
• Slightly reduces the complexity and duration of our build-process
• Moderately reduces the complexity and duration of our test process (by removing the jest transformer)
• Removes ambiguity of local variables named global and the ambient global global.

I am open to discussing any drawbacks this might have! So far I don't really see any besides changes in bundle size.

Resolves #5925

[github-actions]

size-limit report 📦

Path	Size
@sentry/browser - ES5 CDN Bundle (gzipped + minified)	19.41 KB (-0.01% 🔽)
@sentry/browser - ES5 CDN Bundle (minified)	60.09 KB (0%)
@sentry/browser - ES6 CDN Bundle (gzipped + minified)	18.03 KB (-0.08% 🔽)
@sentry/browser - ES6 CDN Bundle (minified)	53.44 KB (+0.88% 🔺)
@sentry/browser - Webpack (gzipped + minified)	19.79 KB (+0.16% 🔺)
@sentry/browser - Webpack (minified)	64.93 KB (+1.01% 🔺)
@sentry/react - Webpack (gzipped + minified)	19.81 KB (+0.14% 🔺)
@sentry/nextjs Client - Webpack (gzipped + minified)	45.53 KB (+0.01% 🔺)
@sentry/browser + @sentry/tracing - ES5 CDN Bundle (gzipped + minified)	25.86 KB (-0.01% 🔽)
@sentry/browser + @sentry/tracing - ES6 CDN Bundle (gzipped + minified)	24.29 KB (+0.01% 🔺)

[Lms24]

I was initially more in favour of changing const->var to const->let which would have solved at least the two Vue-related issues we've talked about. However, seeing that there are a lot of advantages in removing this plugin entirely, I think we can live with the (as you said negligible) bundle size impact.
Also, great that this simplifies our testing setup!