Protect VKalVrt against negative track covariance

Introduce explicit protection against track covariance matrix with negative eigenvalues returned by AtlasExtrapolator. Presumably it happens due to material effects in track extrapolation when the vertex fitter tries to fit a fake vertex candidate.

This merge request affects 1 package:

• Tracking/TrkVertexFitter/TrkVKalVrtCore

added 21.2 Tracking review-pending-level-1 labels

added alsoTargeting:21.0 label

CI Result SUCCESS

	AnalysisBase	AnalysisTop	AthAnalysis	AthDerivation
externals
cmake
make
required tests
optional tests

Full details available at NICOS MR-24469-2019-06-26-20-30
AnalysisBase: number of compilation errors 0, warnings 0
AnalysisTop: number of compilation errors 0, warnings 0
AthAnalysis: number of compilation errors 0, warnings 0
AthDerivation: number of compilation errors 0, warnings 0
For experts only: Jenkins output [CI-MERGE-REQUEST 39698]

Changes look fine to me. @krumnack, please, can you merge?

Cheers, Miguel (AR)

added review-approved label and removed review-pending-level-1 label

mentioned in commit 77a3744c

merged

added sweep:done label

mentioned in merge request !24475 (merged)

mentioned in commit 49d8ce64

Sweep summary
successful:

• 21.0

(see https://aiatlas154.cern.ch/jenkins/job/auto_git_sweeps/826/console for details)

mentioned in merge request !24341 (merged)

mentioned in commit e1c434cd

mentioned in merge request !24523 (merged)

mentioned in commit mmuskinj/athena@e745974a

mentioned in merge request !24614 (merged)

mentioned in commit d9dd5a6d