Force the use of jdk8+ version of Bouncy Castle library

[fab-10]

PR description

This avoid that any dependency can include the jdk5 version in the distribution.

Implement the solution described here https://docs.gradle.org/7.6/userguide/dependency_capability_conflict.html

Had to inline the xor function since it is now deprecated in newer versions of Bouncy Castle.

Fixed Issue(s)

[github-actions]

• I thought about documentation and added the doc-change-required label to this PR if updates are required.
• I thought about the changelog and included a changelog update if required.
• If my PR includes database changes (e.g. KeyValueSegmentIdentifier) I have thought about compatibility and performed forwards and backwards compatibility tests

[fab-10]

@ahamlat did some new test with the last version of Bouncy Castle on a Standard_D4s_v5 instance, and I found that the differences between the distribution are very small, as a side node running on newer jdk gives a performance boost.
Could you validate the results?

Bouncy Castle 1.76

jdk15t018 (java19)

Benchmark                    (SIZE)  Mode  Cnt    Score   Error  Units
KeccakBench.benchmarkKeccak      32  avgt   50  540.001 ± 4.328  ns/op

jdk18on (java19)

Benchmark                    (SIZE)  Mode  Cnt    Score   Error  Units
KeccakBench.benchmarkKeccak      32  avgt   50  544.367 ± 3.604  ns/op

jdk15t018 (java17)

Benchmark                    (SIZE)  Mode  Cnt    Score   Error  Units
KeccakBench.benchmarkKeccak      32  avgt   50  608.889 ± 7.984  ns/op

jdk18on (java17)

Benchmark                    (SIZE)  Mode  Cnt    Score    Error  Units
KeccakBench.benchmarkKeccak      32  avgt   50  620.537 ± 14.868  ns/op

[garyschulte]

This is JDK 1.8+ right? not jdk 18+

Looks good to me, but it would be wise to get 👀 from @ahamlat

[fab-10]

This is JDK 1.8+ right? not jdk 18+

Correct, let me amend the title

[ahamlat]

I was able to reproduce similar numbers regarding JDK 17 and JDK 19. JDK 21 has even better performance for both versions and I didn't notice a difference in performance.

These are some results on Mac M1 Aarch64

Java 21 - jdk15to18 

Benchmark                                         (SIZE)  Mode  Cnt    Score   Error  Units
HashBenchmarking.benchmarkKeccakVersionjdk15to18      32  avgt   50  290.137 ± 2.660  ns/op


Java 21 - jdk18on
Benchmark                                       (SIZE)  Mode  Cnt    Score   Error  Units
HashBenchmarking.benchmarkKeccakVersionjdk18on      32  avgt   50  290.327 ± 2.950  ns/op