fix: address argument parsing silent drops - #817

What does this PR do?:

Hardens Arguments::parse so malformed or unrecognized profiler options are no longer silently dropped. Known options with an invalid value now raise a hard IllegalArgumentException (matching the existing fjmethodid convention); unrecognized top-level tokens and referencechains sub-option keys are now captured and reported as non-fatal warnings via Log::warn, instead of vanishing with no trace.

Motivation:

Several option parsers grouped their default case with the "enable" branch, so a typo silently enabled a feature instead of failing. For example, mcleanup=xyz (a typo) previously fell through to the same branch as mcleanup=true and silently enabled method cleanup. The user would get a profiler running in a state they never asked for, with zero indication anything was wrong. Other options (cstack, wallsampler, jvmtistacks, lightweight, referencechains) had the same class of bug in the opposite direction, silently coercing typos to a default/disabled state. Unknown top-level tokens and referencechains sub-keys were also dropped entirely, even from the (currently non-functional) warning log path.

Additional Notes:

• Added a shared parseBoolOption helper for the y/yes/t/true/1 vs n/no/f/false/0 convention used across these options, replacing several duplicated switch(value[0]) blocks that matched on the first character only (another latent bug: mcleanup=tweq used to pass as true).
• _unknown_arg (single slot) replaced with _unknown_args (vector), so every unmatched token is reported, not just the first.
• Fixed a call-order bug in execute0 (javaApi.cpp) where Log::open(args) ran after the error check/early-return, so warnings were swallowed whenever a hard error also fired in the same command string.

How to test the change?:

new/updated TEST_F(ArgumentsTest, ...) cases in arguments_ut.cpp, covering every hardened option (both the new hard-error and the still-valid-value cases) plus unknown-token capture. Also updated remoteargs_ut.cpp's RemoteSymbolicationInvalidValue test, which previously asserted the silent-enable bug as correct behavior. It now asserts remotesym=invalid is a parse error.

For Datadog employees:

• If this PR touches code that signs or publishes builds or packages, or handles
  credentials of any kind, I've requested a security review (run the dd:platform-security-review
  skill, or file a request via the PSEC review form).
  bewaire also runs automatically on every PR.
• This PR doesn't touch any of that.
• JIRA: [JIRA-16048]

Unsure? Have a question? Request a review!