support bundle ereport directories must be namespaced by part

[hawkw]

In #8739, the code for collecting ereports into support bundles stores the ereports in directories for each serial number from which an ereport was received. Unfortunately, @rmustacc reminds me that I had failed to consider that 0XV1 serial numbers are only unique within the namespace of a particular part, and not globally --- so (for example) a switch and a compute sled may have colliding serials. This means that the current code could incorrectly group ereports reported by two totally different devices. While the ereport JSON files do contain additional information that includes the part number and other identifying information, and restart IDs are additionally capable of distinguishing between reporters, putting ereports from two different systems within the same directory still has the potential to be quite misleading.

Thus, we should probably change this to namespace the ereport directory by which part generated an ereport in addition to the serial number, so that two different devices with colliding serials go in different directories.

[rmustacc]

Just a small note, the 0XV1 bit generally refers to barcode scheme and is true for both 0XV1/0XV2 barcodes. The first version of serial numbers we had (BRMXXXXXXX) appears in both. And a more general note depending on whether we're only focused on sleds / reporters here, we have no guarantees that third-party serials will not overlap with ours (and should assume they will).

[hawkw]

Thanks @rmustacc, I had conflated the two barcode versions and the two serial scheme versions --- is it correct to just refer to the second version of the serial scheme as "version 2" then?

Regarding your other point:

And a more general note depending on whether we're only focused on sleds / reporters here, we have no guarantees that third-party serials will not overlap with ours (and should assume they will).

Yeah, that's useful to bear in mind. Fortunately, the serial numbers used here are always going to be Oxide-issued, it's going to be the serial number of the sled/switch/PSC that reported the ereport. Serials for third-party parts that were involved in the ereport will be in the JSON but not in the path.