p_ext_0_none fails with:
"runtime error: shift exponent 127 is too large for 32-bit type 'int'"

concurrent_hash_map_layout_0_none fails with:
"runtime error: reference binding to null pointer of type ..."

Cast to unsigned was in wrong place. We casted result of assignemnt
which already overflowed which resulted in the follwowing error in
multiple radix tests under UBSAN:

"runtime error: signed integer overflow: 5 + 2147483647 cannot be
represented in type 'int'"

Previously it was default-initialized to nullptr in copy ctor
from non-const iterator which resulted in the following error
for multiple radix tests under UBSAN:

"runtime error: member call on null pointer of type ..."

Original implementation resulted in the following error in multiple
radix tests under UBSAN:

"runtime error: reference binding to misaligned address 0x7fc4f7bc08dc
for type 'const double', which requires 8 byte alignment"

Radix tree concurrent fix

mpsc_queue: improvements needed by multi-tier engine

in radix_tree_iterator::operator++/-- <-> internal_bound. The recursion
caused stack overflow errors, e.g. : https://github.com/pmem/libpmemobj-cpp/issues/1091

Fix Clang detection

CI: add new workflow with more jobs

- use CMAKE_CXX_COMPILER_VERSION/CMAKE_C[XX]_COMPILER_ID for
  setting compiler conditions
- fixes #1090

by renaming 'other_OSes.yml' to 'nigthly.yml' and reordering some jobs.

executing USAN, UBSAN and valgrind tests using upstream valgrind's package.

vector: Fix undefined behaviour on realloc

in try_consume_batch. In case producers constantly produce
new data, try_consume_batch will be called over and over again
leading to stack overflow.

Mpsc queue fix

On tests listed bellow the reserve method is being called before any
allocation on the object therefore realloc being called without
any previous allocation. Inside realloc _data is being used with the
operator '[]', as it is nullptr at that moment it's an undefined
behaviour.

This patch simply returns a call to alloc if _data is nullptr.

This tests fails on PowerPC with Segmentation Fault because of this
issue:
segment_vector_array_expsize_assign_exceptions_oom_0_none
segment_vector_array_expsize_assign_exceptions_oom_0_memcheck
segment_vector_array_expsize_capacity_exceptions_oom_0_none
segment_vector_array_expsize_capacity_exceptions_oom_0_memcheck
segment_vector_array_expsize_modifiers_exceptions_oom_0_none
segment_vector_array_expsize_modifiers_exceptions_oom_0_memcheck
segment_vector_vector_expsize_assign_exceptions_oom_0_none
segment_vector_vector_expsize_assign_exceptions_oom_0_memcheck
segment_vector_vector_expsize_capacity_exceptions_oom_0_none
segment_vector_vector_expsize_capacity_exceptions_oom_0_memcheck
segment_vector_vector_expsize_modifiers_exceptions_oom_0_none
segment_vector_vector_expsize_modifiers_exceptions_oom_0_memcheck
segment_vector_vector_fixedsize_assign_exceptions_oom_0_none
segment_vector_vector_fixedsize_assign_exceptions_oom_0_memcheck

Signed-off-by: Lucas A. M. Magalhães <lamm@linux.ibm.com>

PMREORDER maker was not ended correctly

Change `produce_size` type to avoid sign-compare error

mpsc_queue: extend queue to support arbitrary sized values

It needs to be further debugged

- rename try_consume to try_consume_batch
- rename read_accessor to batch_type
- clear cachelines only after user callback executes:
  this allows iterating over the data arbitrary number of times
  (including 0)
- call user callback within a transaction to simplify logic

- remove recover() method (try_consume can now be used)
- store written offset on pmem so that we can recover
  consumer/producer offsets on restart
- call user callback twice in try_consume if wraparound
  happens

And remove size parameter from mpsc_queue ctor - rely on pmem_log_type size.

Fix windows CI build

we want to enable builds, but current queue implementation
is problematic on Windows - it will be fixed along other queue changes.

Radix: concurrent iteration

Iterator operator++/--:
 Removing nodes/leaves which are pointed to be iterators might
 result in operator++/operator-- failures (since nodes/leaves are
 detached from the tree find_leaf/next_leaf functions will not
 find a proper entry).

 Fix this by adding a fallback path which just searches from the top
 (using lower_bound/upper_bound).

internal_bound:
 Do not traverse the tree twice. Instead save the path from first
 traversal to a vector and just iterate over the vector. Before
 returning iterator to the user, we can check if the path changed
 (if any node on the path was modified). If so, the operation needs
 to be retried.

Revert "gha: temporarily disable codecov"