Fix wrong isPrime results, an EOF loop in the sprint4 map reader and a mis-decoded morse colon

[krotname]

Summary

• kyu6.Prime.isPrime returned true for composite numbers. It called BigInteger.isProbablePrime((int) Math.log(num) + 1), so the certainty was derived from the number itself: for 9 the certainty is 3, i.e. Miller-Rabin was allowed to be wrong with probability up to 1/2^3. isPrime(9) reported true on roughly 2% of runs — a non-deterministic kata answer. Replaced with deterministic trial division (num < 2 → false, even numbers handled, odd divisors up to sqrt(num)), which is also faster for int.
• algorithms.sprint4.Map.Reader replayed its buffer after EOF. read() refilled only when ptr == len; at EOF len becomes -1 with ptr == 0, so the condition was false and the method kept returning stale bytes from the previous buffer (and eventually ran past its end). ptr >= len makes EOF sticky.
• nextCommand() could spin forever. Its whitespace skip loop had no EOF check, unlike nextInt(), so a truncated input (n larger than the number of commands actually supplied) left it looping on -1 forever. It now throws IOException("Unexpected end of input"), which main already treats as rejected input.
• Morse ---... decoded to ;. ---... is a colon; the semicolon -.-.-. was already mapped correctly one line above, so : could never be decoded. Also removed two leftover debug entries in the reverse table ("a" → "b", "c" → "d").
• The diamond kata emitted CRLF on Windows. GiveMeDiamond.print joined rows with System.lineSeparator(); the kata expects \n. TowerBuilder.show() was left alone — it is a display helper, not the kata answer.
• dependabot.yml had two entries targeting master, a branch that does not exist in this repository (default is main), so they could only ever error. Removed; the main entries for maven and github-actions stay.

Not changed (needs a decision)

• algorithms/sprint0/Utils.printList swallows IOException per element. There is an explicit test (UtilsTest.printListSwallowsWriterFailures) asserting exactly that, so the behaviour is deliberate; changing it means changing the signature and the test.
• algorithms/sprint6/DorogayaSet uses -1 as a "not found" sentinel, which collides with a legal answer. Fixing it changes the public method contract.
• Surefire only includes **/*Test.java, **/*Tests.java, **/*TestCase.java and **/*IT.java, so the tag-based suites in src/test/java/quality/*Suite.java never run in the build. Wiring them in would duplicate every test run — a decision about how the suites are meant to be used.

Test plan

• mvn -B -ntp verify (JDK 21) — BUILD SUCCESS, Tests run: 684, Failures: 0, Errors: 0, Skipped: 0, including checkstyle, PMD, SpotBugs and the 70% JaCoCo gates.
• New assertions: PrimeTest now checks primes (2, 3, 5, 7, 11, 13, 101, 7919, 2147483647) and non-primes (-7, -1, 0, 1, 4, 9, 15, 25, 49, 121, 7917, 2147483645); the old code failed on 9 intermittently.
• New MapTest.readerReportsEndOfInputInsteadOfReplayingTheBuffer feeds "g 7" on System.in and asserts that the next nextCommand()/nextInt() throw instead of replaying the buffer; it carries @Timeout(10) so the old infinite loop would fail rather than hang the build.
• MorseCodeDecoderTest gained the case "---... -.-.-." → ":;".

🤖 Generated with Claude Code