fusee_cpp: tweaks, now completes SecureInitialize on hardware

fusee-cpp: a little more init in SecureInitialize

fusee-cpp: finish SecureInitialize

fusee-cpp: implement SecureInitialize besides InitializeClock()

fusee-cpp: sketch out remainder of secure initialize

fusee-cpp: Implement mbist workaround

fusee-cpp: setup exception handlers during crt0

fusee-cpp: add basic structural stubs

ams: bump version to 0.20.1

fusee: fix MSELECT clkrst values

fusee: restore DRAM to 204MHz before starting main cpu on Mariko (closes #1600). ...

Nintendo's Mariko tables result in trained frequency of 1599999 instead of 1600000.

PCV checks for rate == 1600000 exactly, when doing EMC init.

Thus EMC init does not succeed if we are trained to 1600000.

PCV has a fudge factor of 1000 used in SetEmcDvfsFreq, but this is not used in InitEmcDvfs.

This failure means that PCV cannot change rate back to 204MHz before sleep, and then after
wake extremely degraded performance is observed.

Restoring DRAM to 204MHz before boot causes EMC init to succeed/fixes performance degradation.

git subrepo push libraries ...

subrepo:
  subdir:   "libraries"
  merged:   "87a1de0b"
upstream:
  origin:   "https://github.com/Atmosphere-NX/Atmosphere-libs"
  branch:   "master"
  commit:   "87a1de0b"
git-subrepo:
  version:  "0.4.1"
  origin:   "???"
  commit:   "???"

ams: bump to 0.2.0 ...

NOTE: Release will not occur until hekate is ready to handle new sept-less release format.

docs: wipe sept from the docs

ams: replace sept with tsec firmware (#1594) ...

* ams: replace sept with tsec firmware

This replaces sept with a custom tsec key derivation firmware.

NOTE: This does not use any TSEC exploits whatsoever; it is a well-signed
TSEC binary assembled with envyas and signed with the real cauth key.

For more details, contact SciresM#0524.

* fusee: only set SBK if it's readable

kern: improve some debug output

kern: fix same is_current check in GetThreadContext

kern: fix check for thread-is-current in KDebugBase::SetThreadContext

mtc: implement memory training for mariko (#1593) ...

* mtc: implement memory training for mariko

* mtc: fix apply_periodic_compensation_trimmer, train_wr_vref results

* mtc: fix clktree calculations

kern: remove firmare-specific SignalAndModify impl ...

Nintendo made this breaking change because there was zero official software relying on it,
so it's safe for us to make the same change.

dmnt: fix opcode decoding (closes #1575)

kern: fix hardware watchpoint detection

kern: fix inverted condition in context breakpoint validation

dmnt: add theoretical 'else' support to cheat engine vm conditionals

dmnt: various cheat changes/suggestions that have been cooking a while

exo/daybreak: advertise (and check against) supported hos version

Updates to Cheats documentation (#1568) ...

* docs: update cheats doc to clarify Code Type 5 encoding

This change removes references to the "M" nibble for the "Register Address" encoding of Code Type 5 whis is not used in this mode.

In the dword block the "M" has been replaced with a "0"

* docs: update cheats doc to clarify register usage on Code Type 5

This adds additional clarification for "Register Address Encoding" mode of Code Type 5 that the "R" nibble reflects both the destination as well as the base memory address.

* docs: update cheats doc to make Code Types consistent

Code Types are now consistently written in hex notation.

exo: fix SE aes-ctr usage in GetSecureData

kern: since 10.0.0, KDebug::GetThreadContext always returns X0-X7

kern: fix incorrect waiter management in KProcess::ReleaseUserException