Implement travis_retry to reduce the chance that network errors cause CI builds to fail

[johnbillion]

Travis CI has a travis_retry shell function that can be prefixed onto installation commands and cause them to retry up to three times in the event of a non-zero response.

Let's implement this for all commands called during before_install and before_script that touch the network, for example the svn and composer commands. This will reduce the chance that a network issue will cause the build to fail.

Ref: ​https://docs.travis-ci.com/user/common-build-problems/#timeouts-installing-dependencies

[johnbillion]

@netweb Any thoughts on this? I can't see a downside myself.

[netweb]

Replying to johnbillion:

@netweb Any thoughts on this? I can't see a downside myself.

Nope, go for it

[johnbillion]

In 43645:

Build/Test Tools: Implement the automatic retry functionality that Travis CI provides.

This functionality will cause the command to be retried up to three times in case of a non-zero return value. Implementing it on commands that perform network requests means that intermittent network failures are less likely to cause a build to fail, as they'll be retried up to three times.

See #44858

[pento]

Reopen for merge to the 5.0 branch.

[pento]

In 43724:

Build/Test Tools: Implement the automatic retry functionality that Travis CI provides.

This functionality will cause the command to be retried up to three times in case of a non-zero return value. Implementing it on commands that perform network requests means that intermittent network failures are less likely to cause a build to fail, as they'll be retried up to three times.

Merges [43645] to the 5.0 branch.

Props johnbillion.
Fixes #44858.

[pento]

In 43725:

Build/Test Tools: Remove an extra line incorrectly included in [43724].

See #44858.