What

Parse the NC RTX PRO 6000 BSE v6 GPU VM sizes, e.g. Standard_NC288ds_xl_RTXPRO6000BSE_v6.

Today these fail with could not parse VM size ... because the size scheme does not handle two new elements in the name:

1. the xl_ size descriptor that precedes the accelerator, and
2. an accelerator name with trailing letters (RTXPRO6000BSE), while the accelerator group only matched <letters><digits> ([A-Z]+[0-9]+).

These sizes are public and documented: https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/gpu-accelerated/nc-rtxpro6000-bse-v6-series

Change

Extend the accelerator capture group:

(?:([A-Z]+[0-9]+)_?)?   ->   (?:(xl_[A-Z]+[0-9]+[A-Z]*|[A-Z]+[0-9]+)_?)?

• The trailing-letters form is gated behind the literal xl_ so existing accelerators (MI300X, H100, A100, T4, …) keep matching their original/dedicated groups and their series flags (MI300Series, H100Series) are unchanged. (An ungated [A-Z]* would have swallowed MI300X into the accelerator group and flipped MI300Series.)
• The xl_ descriptor is stripped so AcceleratorType reflects the accelerator name only (RTXPRO6000BSE).

Applied to both the root module and v2/.

Tests

• Added Test_GetVMSize cases for the ds and lds variants (NC288ds_xl_RTXPRO6000BSE_v6, NC24lds_xl_RTXPRO6000BSE_v6).
• TestParseVMSize over the generated real-SKU corpus passes with 0 failures (no regressions), in both modules.
• gofmt, go test ./... green for root and v2.

Note

The xl token is not yet listed in the VM sizes naming conventions doc; the series page above is the current public reference.