Add keepTrailingNewline option to LegacyOverrides to match Python Jinja2 default behaviour

[jmoraleda]

Description:

By default, Python Jinja2 strips a single trailing newline from rendered output (keep_trailing_newline=False). Jinjava has historically preserved trailing newlines, causing a subtle but consistent difference when porting templates from Python Jinja2.

This PR adds a keepTrailingNewline flag to LegacyOverrides that allows users to opt into Python-compatible behaviour.

Changes:

• LegacyOverrides: adds isKeepTrailingNewline(), defaulting to true (preserve existing Jinjava behaviour — no impact on current users).
• JinjavaInterpreter: applies the strip in stripTrailingNewlineIfNeeded(), called at the normal return points of the private render() method. Early-exit error paths (OutputTooBigException, CollectionTooBigException) are intentionally unaffected.
• TrailingNewlineTest: covers both flag values, the NONE/THREE_POINT_0/ALL presets, and edge cases (no trailing newline, multiple trailing newlines, rendered expressions).

Usage:

To opt into Python-compatible behaviour:

JinjavaConfig config = JinjavaConfig.newBuilder()
    .withLegacyOverrides(
        LegacyOverrides.newBuilder()
            .withKeepTrailingNewline(false)
            .build()
    )
    .build();

Compatibility note:

keepTrailingNewline is set to true (legacy behaviour) in THREE_POINT_0 to avoid breaking existing tests. It is set to false in ALL. Maintainers may wish to move it to false in THREE_POINT_0 as well if they consider stripping the trailing newline to be part of the canonical Python-compatible behaviour set — the change is intentionally conservative here to leave that decision to the maintainers.

[jasmith-hs]

Thanks @jmoraleda for the PR. Could you change it so that the default is false like the other legacy override options and rename it something like stripTrailingNewlines?
Also, as Jinjava 3.0 has not been published yet, so this option can be turned on for 3.0 so that the default behaviour of Jinjava 3.0 is as close to Jinja as possible

[jmoraleda]

Thank you @jasmith-hs :

1. I am ok changing it to stripTrailingNewlines and reversing the logic. But since in Python jinja they use

   

   Can you please confirm you still want to change the name and reverse the logic?

2. I am happy to include this change in THREE_POINT_0. Since THREE_POINT_0 is the default used for testing, that breaks some of the existing tests. For tests that break, do you prefer me to change the expected output of the tests to match THREE_POINT_0 default behavior, or do you prefer to modify the test to use a modified configuration that preserves the older behavior?

3. Should I also modify fix: treat backslash as escape character only inside quoted strings, matching Jinja2 behaviour (fixes #1304) #1306 to add that feature to THREE_POINT_0 configuration?

[jasmith-hs]

Ah, thanks @jmoraleda, since it's configurable in Jinja, I believe that it would be best to have the option added to the top-level JinjavaConfig. We an have this be defaulting to false to match Jinja behaviour in the Jinjava 3.0 release since it's a major version bump and we can have breaking changes. (I may do a final 2.8.4 release that includes these some new changes such as these couple that you're PR-ing and have the 2.8.4 version use non-breaking defaults)

2. Whatever is easier, really. Slight preference for setting the keep trailing slash on

3. Yes, I'd recommend setting it to true in THREE_POINT_0, which will be the culmination of everything that currently exists turned on at the time of 3.0.0 release, but then after release, it will not have any breaking changes. Since 3.0.0 isn't released yet, the THREE_POINT_0 option can have the correct (breaking) functionality turned on