The release-size budget only reports that the backupstrategy-controller
release grew too large, not which field did it, and three strategies
once inlined the upstream pod schema before anyone noticed. Check the
generated strategy CRDs for an inlined PodSpec and name the property
path, with the markers that fix it.

Assisted-by: LLM
Signed-off-by: Myasnikov Daniil <[email protected]>
(cherry picked from commit 513b508)

A chart can grow past the 1 MiB Helm release Secret without any test
noticing: it renders, lints and installs, and only the upgrade whose
manifest is large enough fails, on a live cluster, with a release that
stays failed. backupstrategy-controller sat under 1 KB from the limit
before it crossed it.

Estimate the stored size the way Helm encodes it (release JSON with the
chart's templates and files, gzip, base64) and fail any chart under
packages/ above 90 percent of the limit, plus the largest revision of
backupstrategy-controller with every default Strategy rendered. The
estimate counts files a .helmignore names, because the production path
stores them.

A size measured without its manifest is a lower bound that passes for
the wrong reason, so every chart must render offline except an explicit
list of charts that need cluster state, and the worst-case render must
succeed and contain the default strategies before it is measured.

Assisted-by: LLM
Signed-off-by: Myasnikov Daniil <[email protected]>
(cherry picked from commit 7ad7d39)