build: use JSpecify @Nullable instead of JSR305

[nielspardon]

JSR305 has been dormant for years and was never finalized. It would be better to use alternative approaches like the checkerframework which is e.g. what Apache Calcite is doing.

[nielspardon]

looks like a transient build error due to our faulty Java version config. should be fixed by #443 or #433

[nielspardon]

rebased on latest main and the build issue is gone

[bestbeforetoday]

Based on a little searching, I am not sure that checkerframework is something we should be buying in to:

• Please consider a different library other than Checker-Framework for JSR305 replacement google/guava#3031
• Migrate off of jsr305 google/guava#2960
• Adoption of JSpecify annotations in Guava jspecify/jspecify#239

JSpecify seems to have a lot of industry backing (Google, Jetbrains, Meta, Microsoft, Oracle, PMD, ...) so maybe that is a better direction?

[nielspardon]

Based on a little searching, I am not sure that checkerframework is something we should be buying in to:

I don't mind which alternative we use we just shouldn't use JSR305 in my opinion.

[nielspardon]

I rebased on current main and pushed a commit switching to JSpecify

[bestbeforetoday]

This looks good to me but really needs the blessing of @vbarua on the preferred nullability annotation library.

[vbarua]

I actually don't have strong preferences here. As @/bestbeforetoday points out, JSpecify seems to have broad industry support so I'm happy to switch to it.

[vbarua]

LGTM