Remove delta.experimental.ignore-checkpoint-write-failures config

[homar]

Description

In case of failure while writing checkpoint file error is logged.

Non-technical explanation

Release notes

() This is not user-visible and no release notes are required.
( ) Release notes are required, please propose a release note for me.
(x) Release notes are required, with the following suggested text:

# delta
* Failure while writing checkpoint will result in a try to undo the operation. ({issue}`issuenumber`)

[findepi]

writeCheckpointIfNeeded is called after the write has been "committed", i.e. the transaction log exists.
Let's assume it's an INSERT query. If we now fail it, it can be retried by the calling application, leading to duplicate rows.

IMO we should log unconditionally

@alexjo2144 wdyt?

[losipiuk]

What loggin buys us? If so I would make exception message (for which it is more probable that someone would actually read it) more explanatory (that data could have been (or have been) already persisted in the table even though query failed).

[losipiuk]

If write was committed already then created files will be remained, otherwise they will be cleaned up.

Does "otherwise" ever happen? Look it does not, which makes RN text misleading iMO.

[alexjo2144]

Add a @DefunctConfig annotation

[findepi]

Let's capture the reasoning from our conversation, why we don't fail here

[homar]

Comment added