feat: Add sourcemaps inject command

[loewenheim]

This implements #1467.

It adds a new subcommand inject PATH to the sourcemaps command that works as follows:

For each .js file file.js in PATH,

1. If file.js contains a //# sentryDebugId //# debugId comment, skip it (it has already been processed).
2. If file.js does not contain a //# sourceMappingURL=<sourcemap_path> comment, skip it (without a sourcemap there's nothing to do).
3. Check if sourcemap_path exists. If not, emit a warning and skip file.js.
4. Check if the sourcemap at sourcemap_path contains a debug id under the debugID x_sentry_debug_iddebug_id key. If so, return it, otherwise insert a fresh debug id under that key and return that.
5. Append a JS code snippet and a //# sentryDebugId=__SENTRY_DEBUG_ID__ comment to file.js, where __SENTRY_DEBUG_ID__ is replaced with the debug id returned in step 4.

This procedure differs slightly from the description in #1467: it checks first whether the sourcemap already contains a debug id instead of unconditionally generating a new one.

[kamilogorek]

They suppose to be terminal line-break specific to clap, but for some reason you say they stopped working? 🤔

https://github.com/clap-rs/clap/blob/ad5d67623a89415c7f5f7e1f7a47bf5abdfd0b6c/src/output/help_template.rs#L1018

[loewenheim]

The placeholders have apparently been in the send_envelope-help test case since the beginning.

https://github.com/getsentry/sentry-cli/blame/master/tests/integration/_cases/send_envelope/send_envelope-help.trycmd

[kamilogorek]

I'll review this next week, as I'm not even sure what we exactly want to achieve here tbh.

[mitsuhiko]

I would vote to change x_sentry_debug_id to debug_id and sentryDebugId to just debugId. While right now we're clearly the main user here, I'm not sure adding a vendor prefix is doing anything in particular here. Not sure if anyone has strong opinions here.

[loewenheim]

This now uses sourcemap-related functions that were added in symbolic 12.0.0. I couldn't find a good way to use discover_sourcemap_embedded_debug_id because we need to write to the sourcemap file anyway.

[mitsuhiko]

@loewenheim I think fixup_js_file should be a no-op if discover_sourcemap_embedded_debug_id returns Some. Otherwise that run is not idempotent and we add the snippet twice.

[loewenheim]

But if the snippet was already added, the file should have a //# debugId comment and been skipped in the loop anyway. Or am I misunderstanding you?

[loewenheim]

@lforst has convinced me that backing up the files before modifying them is the wrong thing to do, for two reasons:

1. The files are easily restored from version control or regenerated if necessary.
2. sentry-cli spamming .bak files into your directory would be very annoying.